--- lustre-1.8.3.orig/ldiskfs/kernel_patches/series/ldiskfs-2.6.18-debian.series +++ lustre-1.8.3/ldiskfs/kernel_patches/series/ldiskfs-2.6.18-debian.series @@ -0,0 +1,20 @@ +ext3-wantedi-2.6-rhel4.patch +iopen-2.6-fc5.patch +ext3-map_inode_page-2.6.18.patch +export-ext3-2.6-rhel4.patch +ext3-include-fixes-2.6-rhel4.patch +ext3-extents-2.6.18-vanilla.patch +ext3-mballoc3-core.patch +ext3-mballoc3-2.6.18.patch +ext3-nlinks-2.6.9.patch +ext3-ialloc-2.6.patch +ext3-remove-cond_resched-calls-2.6.12.patch +ext3-filterdata-sles10.patch +ext3-16tb-overflow-fixes.patch +ext3-uninit-2.6.18.patch +ext3-nanosecond-2.6.18-vanilla.patch +ext3-inode-version-2.6.18-vanilla.patch +ext3-ea-expand-lose-block.patch +ext3-mmp-2.6.18-vanilla.patch +ext3-fiemap-2.6.18-vanilla.patch +ext3-lookup-dotdot-2.6.9.patch --- lustre-1.8.3.orig/lustre/kernel_patches/patches/quota-support-64-bit-quota-format-2.6.27-vanilla.patch +++ lustre-1.8.3/lustre/kernel_patches/patches/quota-support-64-bit-quota-format-2.6.27-vanilla.patch @@ -0,0 +1,282 @@ +From: Jan Kara + +Implement conversion functions for new version (version 1) of quota format +which supports 64-bit block and inode limits and 64-bit inode usage. The +original implementation has been written by Andrew Perepechko. + +Signed-off-by: Andrew Perepechko +Signed-off-by: Jan Kara +Signed-off-by: Andrew Morton +--- + + fs/quota_v2.c | 140 ++++++++++++++++++++++++++++++++++++---------- + fs/quotaio_v2.h | 26 ++++++-- + 2 files changed, 132 insertions(+), 34 deletions(-) + +Index: linux-2.6.27.36/fs/quota_v2.c +=================================================================== +--- linux-2.6.27.36.orig/fs/quota_v2.c 2009-10-09 17:00:48.000000000 +0200 ++++ linux-2.6.27.36/fs/quota_v2.c 2009-10-09 17:03:35.000000000 +0200 +@@ -23,14 +23,24 @@ + + #define __QUOTA_V2_PARANOIA + +-static void v2_mem2diskdqb(void *dp, struct dquot *dquot); +-static void v2_disk2memdqb(struct dquot *dquot, void *dp); +-static int v2_is_id(void *dp, struct dquot *dquot); +- +-static struct qtree_fmt_operations v2_qtree_ops = { +- .mem2disk_dqblk = v2_mem2diskdqb, +- .disk2mem_dqblk = v2_disk2memdqb, +- .is_id = v2_is_id, ++static void v2r0_mem2diskdqb(void *dp, struct dquot *dquot); ++static void v2r0_disk2memdqb(struct dquot *dquot, void *dp); ++static int v2r0_is_id(void *dp, struct dquot *dquot); ++ ++static struct qtree_fmt_operations v2r0_qtree_ops = { ++ .mem2disk_dqblk = v2r0_mem2diskdqb, ++ .disk2mem_dqblk = v2r0_disk2memdqb, ++ .is_id = v2r0_is_id, ++}; ++ ++static void v2r1_mem2diskdqb(void *dp, struct dquot *dquot); ++static void v2r1_disk2memdqb(struct dquot *dquot, void *dp); ++static int v2r1_is_id(void *dp, struct dquot *dquot); ++ ++static struct qtree_fmt_operations v2r1_qtree_ops = { ++ .mem2disk_dqblk = v2r1_mem2diskdqb, ++ .disk2mem_dqblk = v2r1_disk2memdqb, ++ .is_id = v2r1_is_id, + }; + + #define QUOTABLOCK_BITS 10 +@@ -46,8 +56,7 @@ + return blocks << QUOTABLOCK_BITS; + } + +-/* Check whether given file is really vfsv0 quotafile */ +-static int v2_check_quota_file(struct super_block *sb, int type) ++static int v2_check_quota_file_header(struct super_block *sb, int type) + { + struct v2_disk_dqheader dqhead; + ssize_t size; +@@ -58,12 +67,20 @@ + if (size != sizeof(struct v2_disk_dqheader)) { + printk("quota_v2: failed read expected=%zd got=%zd\n", + sizeof(struct v2_disk_dqheader), size); +- return 0; ++ return -EIO; + } +- if (le32_to_cpu(dqhead.dqh_magic) != quota_magics[type] || +- le32_to_cpu(dqhead.dqh_version) != quota_versions[type]) +- return 0; +- return 1; ++ if (le32_to_cpu(dqhead.dqh_magic) != quota_magics[type]) ++ return -ENOENT; ++ if (le32_to_cpu(dqhead.dqh_version) > quota_versions[type]) ++ return -EOPNOTSUPP; ++ return le32_to_cpu(dqhead.dqh_version); ++} ++ ++ ++/* Check whether given file is really vfsv0 quotafile */ ++static int v2_check_quota_file(struct super_block *sb, int type) ++{ ++ return v2_check_quota_file_header(sb, type) >= 0; + } + + /* Read information header from quota file */ +@@ -72,7 +89,13 @@ + struct v2_disk_dqinfo dinfo; + struct mem_dqinfo *info = sb_dqinfo(sb, type); + ssize_t size; ++ int version = v2_check_quota_file_header(sb, type); + ++ if (version < 0) { ++ printk(KERN_WARNING "Cannot identify quota file version on " ++ "device %s: %d\n", sb->s_id, version); ++ return -1; ++ } + size = sb->s_op->quota_read(sb, type, (char *)&dinfo, + sizeof(struct v2_disk_dqinfo), V2_DQINFOOFF); + if (size != sizeof(struct v2_disk_dqinfo)) { +@@ -81,8 +104,14 @@ + return -1; + } + /* limits are stored as unsigned 32-bit data */ +- info->dqi_maxblimit = 0xffffffff; +- info->dqi_maxilimit = 0xffffffff; ++ if (version == 0) { ++ /* limits are stored as unsigned 32-bit data */ ++ info->dqi_maxblimit = 0xffffffff; ++ info->dqi_maxilimit = 0xffffffff; ++ } else { ++ info->dqi_maxblimit = 0x7fffffffffffffffULL; ++ info->dqi_maxilimit = 0x7fffffffffffffffULL; ++ } + info->dqi_bgrace = le32_to_cpu(dinfo.dqi_bgrace); + info->dqi_igrace = le32_to_cpu(dinfo.dqi_igrace); + info->dqi_flags = le32_to_cpu(dinfo.dqi_flags); +@@ -94,8 +123,13 @@ + info->u.v2_i.i.dqi_blocksize_bits = V2_DQBLKSIZE_BITS; + info->u.v2_i.i.dqi_usable_bs = 1 << V2_DQBLKSIZE_BITS; + info->u.v2_i.i.dqi_qtree_depth = qtree_depth(&info->u.v2_i.i); +- info->u.v2_i.i.dqi_entry_size = sizeof(struct v2_disk_dqblk); +- info->u.v2_i.i.dqi_ops = &v2_qtree_ops; ++ if (version == 0) { ++ qinfo->dqi_entry_size = sizeof(struct v2r0_disk_dqblk); ++ qinfo->dqi_ops = &v2r0_qtree_ops; ++ } else { ++ qinfo->dqi_entry_size = sizeof(struct v2r1_disk_dqblk); ++ qinfo->dqi_ops = &v2r1_qtree_ops; ++ } + return 0; + } + +@@ -125,9 +159,9 @@ + return 0; + } + +-static void v2_disk2memdqb(struct dquot *dquot, void *dp) ++static void v2r0_disk2memdqb(struct dquot *dquot, void *dp) + { +- struct v2_disk_dqblk *d = dp, empty; ++ struct v2r0_disk_dqblk *d = dp, empty; + struct mem_dqblk *m = &dquot->dq_dqb; + + m->dqb_ihardlimit = le32_to_cpu(d->dqb_ihardlimit); +@@ -139,15 +173,15 @@ + m->dqb_curspace = le64_to_cpu(d->dqb_curspace); + m->dqb_btime = le64_to_cpu(d->dqb_btime); + /* We need to escape back all-zero structure */ +- memset(&empty, 0, sizeof(struct v2_disk_dqblk)); ++ memset(&empty, 0, sizeof(struct v2r0_disk_dqblk)); + empty.dqb_itime = cpu_to_le64(1); +- if (!memcmp(&empty, dp, sizeof(struct v2_disk_dqblk))) ++ if (!memcmp(&empty, dp, sizeof(struct v2r0_disk_dqblk))) + m->dqb_itime = 0; + } + +-static void v2_mem2diskdqb(void *dp, struct dquot *dquot) ++static void v2r0_mem2diskdqb(void *dp, struct dquot *dquot) + { +- struct v2_disk_dqblk *d = dp; ++ struct v2r0_disk_dqblk *d = dp; + struct mem_dqblk *m = &dquot->dq_dqb; + struct qtree_mem_dqinfo *info = + &sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i; +@@ -165,9 +199,60 @@ + d->dqb_itime = cpu_to_le64(1); + } + +-static int v2_is_id(void *dp, struct dquot *dquot) ++static int v2r0_is_id(void *dp, struct dquot *dquot) ++{ ++ struct v2r0_disk_dqblk *d = dp; ++ struct qtree_mem_dqinfo *info = ++ sb_dqinfo(dquot->dq_sb, dquot->dq_type)->dqi_priv; ++ ++ if (qtree_entry_unused(info, dp)) ++ return 0; ++ return le32_to_cpu(d->dqb_id) == dquot->dq_id; ++} ++ ++static void v2r1_disk2memdqb(struct dquot *dquot, void *dp) ++{ ++ struct v2r1_disk_dqblk *d = dp, empty; ++ struct mem_dqblk *m = &dquot->dq_dqb; ++ ++ m->dqb_ihardlimit = le64_to_cpu(d->dqb_ihardlimit); ++ m->dqb_isoftlimit = le64_to_cpu(d->dqb_isoftlimit); ++ m->dqb_curinodes = le64_to_cpu(d->dqb_curinodes); ++ m->dqb_itime = le64_to_cpu(d->dqb_itime); ++ m->dqb_bhardlimit = v2_qbtos(le64_to_cpu(d->dqb_bhardlimit)); ++ m->dqb_bsoftlimit = v2_qbtos(le64_to_cpu(d->dqb_bsoftlimit)); ++ m->dqb_curspace = le64_to_cpu(d->dqb_curspace); ++ m->dqb_btime = le64_to_cpu(d->dqb_btime); ++ /* We need to escape back all-zero structure */ ++ memset(&empty, 0, sizeof(struct v2r1_disk_dqblk)); ++ empty.dqb_itime = cpu_to_le64(1); ++ if (!memcmp(&empty, dp, sizeof(struct v2r1_disk_dqblk))) ++ m->dqb_itime = 0; ++} ++ ++static void v2r1_mem2diskdqb(void *dp, struct dquot *dquot) ++{ ++ struct v2r1_disk_dqblk *d = dp; ++ struct mem_dqblk *m = &dquot->dq_dqb; ++ struct qtree_mem_dqinfo *info = ++ sb_dqinfo(dquot->dq_sb, dquot->dq_type)->dqi_priv; ++ ++ d->dqb_ihardlimit = cpu_to_le64(m->dqb_ihardlimit); ++ d->dqb_isoftlimit = cpu_to_le64(m->dqb_isoftlimit); ++ d->dqb_curinodes = cpu_to_le64(m->dqb_curinodes); ++ d->dqb_itime = cpu_to_le64(m->dqb_itime); ++ d->dqb_bhardlimit = cpu_to_le64(v2_stoqb(m->dqb_bhardlimit)); ++ d->dqb_bsoftlimit = cpu_to_le64(v2_stoqb(m->dqb_bsoftlimit)); ++ d->dqb_curspace = cpu_to_le64(m->dqb_curspace); ++ d->dqb_btime = cpu_to_le64(m->dqb_btime); ++ d->dqb_id = cpu_to_le32(dquot->dq_id); ++ if (qtree_entry_unused(info, dp)) ++ d->dqb_itime = cpu_to_le64(1); ++} ++ ++static int v2r1_is_id(void *dp, struct dquot *dquot) + { +- struct v2_disk_dqblk *d = dp; ++ struct v2r1_disk_dqblk *d = dp; + struct qtree_mem_dqinfo *info = + &sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i; + +Index: linux-2.6.27.36/include/linux/quotaio_v2.h +=================================================================== +--- linux-2.6.27.36.orig/include/linux/quotaio_v2.h 2009-10-09 17:00:48.000000000 +0200 ++++ linux-2.6.27.36/include/linux/quotaio_v2.h 2009-10-09 17:00:56.000000000 +0200 +@@ -17,8 +17,8 @@ + } + + #define V2_INITQVERSIONS {\ +- 0, /* USRQUOTA */\ +- 0 /* GRPQUOTA */\ ++ 1, /* USRQUOTA */\ ++ 1 /* GRPQUOTA */\ + } + + /* First generic header */ +@@ -28,11 +28,11 @@ + }; + + /* +- * The following structure defines the format of the disk quota file +- * (as it appears on disk) - the file is a radix tree whose leaves point +- * to blocks of these structures. ++ * The following structure defines the format of the disk quota file in version ++ * 0 - the file is a radix tree whose leaves point to blocks of these ++ * structures. + */ +-struct v2_disk_dqblk { ++struct v2r0_disk_dqblk { + __le32 dqb_id; /* id this quota applies to */ + __le32 dqb_ihardlimit; /* absolute limit on allocated inodes */ + __le32 dqb_isoftlimit; /* preferred inode limit */ +@@ -44,6 +44,20 @@ + __le64 dqb_itime; /* time limit for excessive inode use */ + }; + ++/* The same structure in quota file version 1 */ ++struct v2r1_disk_dqblk { ++ __le32 dqb_id; /* id this quota applies to */ ++ __le32 dqb_padding; /* padding field */ ++ __le64 dqb_ihardlimit; /* absolute limit on allocated inodes */ ++ __le64 dqb_isoftlimit; /* preferred inode limit */ ++ __le64 dqb_curinodes; /* current # allocated inodes */ ++ __le64 dqb_bhardlimit; /* absolute limit on disk space */ ++ __le64 dqb_bsoftlimit; /* preferred limit on disk space */ ++ __le64 dqb_curspace; /* current space occupied (in bytes) */ ++ __le64 dqb_btime; /* time limit for excessive disk use */ ++ __le64 dqb_itime; /* time limit for excessive inode use */ ++}; ++ + /* Header with type and version specific information */ + struct v2_disk_dqinfo { + __le32 dqi_bgrace; /* Time before block soft limit becomes hard limit */ --- lustre-1.8.3.orig/lustre/kernel_patches/patches/export-2.6.26-vanilla.patch +++ lustre-1.8.3/lustre/kernel_patches/patches/export-2.6.26-vanilla.patch @@ -0,0 +1,24 @@ +Index: linux-source-2.6.26/fs/jbd2/journal.c +=================================================================== +--- linux-source-2.6.26.orig/fs/jbd2/journal.c 2009-10-09 13:49:43.000000000 +0200 ++++ linux-source-2.6.26/fs/jbd2/journal.c 2009-10-09 13:49:50.000000000 +0200 +@@ -460,6 +460,7 @@ + spin_unlock(&journal->j_state_lock); + return ret; + } ++EXPORT_SYMBOL(jbd2_log_start_commit); + + /* + * Force and wait upon a commit if the calling process is not within +Index: linux-source-2.6.26/security/security.c +=================================================================== +--- linux-source-2.6.26.orig/security/security.c 2009-10-09 13:49:43.000000000 +0200 ++++ linux-source-2.6.26/security/security.c 2009-10-09 13:49:50.000000000 +0200 +@@ -68,6 +68,7 @@ + + return 0; + } ++EXPORT_SYMBOL(security_inode_unlink); + + /* Save user chosen LSM */ + static int __init choose_lsm(char *str) --- lustre-1.8.3.orig/lustre/kernel_patches/patches/quota-Split-off-quota-tree-handling-into-a-separate.patch +++ lustre-1.8.3/lustre/kernel_patches/patches/quota-Split-off-quota-tree-handling-into-a-separate.patch @@ -0,0 +1,1593 @@ +From: Jan Kara +References: fate#302681 +Subject: [PATCH 12/28] quota: Split off quota tree handling into a separate file +Patch-mainline: 2.6.29? + +There is going to be a new version of quota format having 64-bit +quota limits and a new quota format for OCFS2. They are both +going to use the same tree structure as VFSv0 quota format. So +split out tree handling into a separate file and make size of +leaf blocks, amount of space usable in each block (needed for +checksumming) and structures contained in them configurable +so that the code can be shared. + +Signed-off-by: Jan Kara +--- + fs/Kconfig | 5 + fs/Makefile | 1 + fs/quota_tree.c | 645 ++++++++++++++++++++++++++++++++++++++++++++ + fs/quota_tree.h | 25 + + fs/quota_v2.c | 598 +++------------------------------------- + fs/quotaio_v2.h | 33 -- + include/linux/dqblk_qtree.h | 56 +++ + include/linux/dqblk_v2.h | 19 - + 8 files changed, 800 insertions(+), 582 deletions(-) + create mode 100644 fs/quota_tree.c + create mode 100644 fs/quota_tree.h + create mode 100644 include/linux/dqblk_qtree.h + +Index: linux-2.6.27.36/fs/Kconfig +=================================================================== +--- linux-2.6.27.36.orig/fs/Kconfig 2009-10-05 17:19:01.000000000 +0200 ++++ linux-2.6.27.36/fs/Kconfig 2009-10-08 16:32:48.000000000 +0200 +@@ -569,6 +569,10 @@ + Note that this behavior is currently deprecated and may go away in + future. Please use notification via netlink socket instead. + ++# Generic support for tree structured quota files. Seleted when needed. ++config QUOTA_TREE ++ tristate ++ + config QFMT_V1 + tristate "Old quota format support" + depends on QUOTA +@@ -580,6 +584,7 @@ + config QFMT_V2 + tristate "Quota format v2 support" + depends on QUOTA ++ select QUOTA_TREE + help + This quota format allows using quotas with 32-bit UIDs/GIDs. If you + need this functionality say Y here. +Index: linux-2.6.27.36/fs/Makefile +=================================================================== +--- linux-2.6.27.36.orig/fs/Makefile 2009-10-05 17:19:01.000000000 +0200 ++++ linux-2.6.27.36/fs/Makefile 2009-10-08 16:32:48.000000000 +0200 +@@ -53,6 +53,7 @@ + obj-$(CONFIG_QUOTA) += dquot.o + obj-$(CONFIG_QFMT_V1) += quota_v1.o + obj-$(CONFIG_QFMT_V2) += quota_v2.o ++obj-$(CONFIG_QUOTA_TREE) += quota_tree.o + obj-$(CONFIG_QUOTACTL) += quota.o + + obj-$(CONFIG_DNOTIFY) += dnotify.o +Index: linux-2.6.27.36/include/linux/quotaio_v2.h +=================================================================== +--- linux-2.6.27.36.orig/include/linux/quotaio_v2.h 2009-10-05 17:19:01.000000000 +0200 ++++ linux-2.6.27.36/include/linux/quotaio_v2.h 2009-10-08 16:32:48.000000000 +0200 +@@ -21,6 +21,12 @@ + 0 /* GRPQUOTA */\ + } + ++/* First generic header */ ++struct v2_disk_dqheader { ++ __le32 dqh_magic; /* Magic number identifying file */ ++ __le32 dqh_version; /* File version */ ++}; ++ + /* + * The following structure defines the format of the disk quota file + * (as it appears on disk) - the file is a radix tree whose leaves point +@@ -38,15 +44,6 @@ + __le64 dqb_itime; /* time limit for excessive inode use */ + }; + +-/* +- * Here are header structures as written on disk and their in-memory copies +- */ +-/* First generic header */ +-struct v2_disk_dqheader { +- __le32 dqh_magic; /* Magic number identifying file */ +- __le32 dqh_version; /* File version */ +-}; +- + /* Header with type and version specific information */ + struct v2_disk_dqinfo { + __le32 dqi_bgrace; /* Time before block soft limit becomes hard limit */ +@@ -57,23 +54,7 @@ + __le32 dqi_free_entry; /* Number of block with at least one free entry */ + }; + +-/* +- * Structure of header of block with quota structures. It is padded to 16 bytes so +- * there will be space for exactly 21 quota-entries in a block +- */ +-struct v2_disk_dqdbheader { +- __le32 dqdh_next_free; /* Number of next block with free entry */ +- __le32 dqdh_prev_free; /* Number of previous block with free entry */ +- __le16 dqdh_entries; /* Number of valid entries in block */ +- __le16 dqdh_pad1; +- __le32 dqdh_pad2; +-}; +- + #define V2_DQINFOOFF sizeof(struct v2_disk_dqheader) /* Offset of info header in file */ +-#define V2_DQBLKSIZE_BITS 10 +-#define V2_DQBLKSIZE (1 << V2_DQBLKSIZE_BITS) /* Size of block with quota structures */ +-#define V2_DQTREEOFF 1 /* Offset of tree in file in blocks */ +-#define V2_DQTREEDEPTH 4 /* Depth of quota tree */ +-#define V2_DQSTRINBLK ((V2_DQBLKSIZE - sizeof(struct v2_disk_dqdbheader)) / sizeof(struct v2_disk_dqblk)) /* Number of entries in one blocks */ ++#define V2_DQBLKSIZE_BITS 10 /* Size of leaf block in tree */ + + #endif /* _LINUX_QUOTAIO_V2_H */ +Index: linux-2.6.27.36/fs/quota_tree.c +=================================================================== +--- /dev/null 1970-01-01 00:00:00.000000000 +0000 ++++ linux-2.6.27.36/fs/quota_tree.c 2009-10-08 16:32:48.000000000 +0200 +@@ -0,0 +1,645 @@ ++/* ++ * vfsv0 quota IO operations on file ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++ ++#include ++ ++#include "quota_tree.h" ++ ++MODULE_AUTHOR("Jan Kara"); ++MODULE_DESCRIPTION("Quota trie support"); ++MODULE_LICENSE("GPL"); ++ ++#define __QUOTA_QT_PARANOIA ++ ++typedef char *dqbuf_t; ++ ++static int get_index(struct qtree_mem_dqinfo *info, qid_t id, int depth) ++{ ++ unsigned int epb = info->dqi_usable_bs >> 2; ++ ++ depth = info->dqi_qtree_depth - depth - 1; ++ while (depth--) ++ id /= epb; ++ return id % epb; ++} ++ ++/* Number of entries in one blocks */ ++static inline int qtree_dqstr_in_blk(struct qtree_mem_dqinfo *info) ++{ ++ return (info->dqi_usable_bs - sizeof(struct qt_disk_dqdbheader)) ++ / info->dqi_entry_size; ++} ++ ++static dqbuf_t getdqbuf(size_t size) ++{ ++ dqbuf_t buf = kmalloc(size, GFP_NOFS); ++ if (!buf) ++ printk(KERN_WARNING "VFS: Not enough memory for quota buffers.\n"); ++ return buf; ++} ++ ++static inline void freedqbuf(dqbuf_t buf) ++{ ++ kfree(buf); ++} ++ ++static inline ssize_t read_blk(struct qtree_mem_dqinfo *info, uint blk, dqbuf_t buf) ++{ ++ struct super_block *sb = info->dqi_sb; ++ ++ memset(buf, 0, info->dqi_usable_bs); ++ return sb->s_op->quota_read(sb, info->dqi_type, (char *)buf, ++ info->dqi_usable_bs, blk << info->dqi_blocksize_bits); ++} ++ ++static inline ssize_t write_blk(struct qtree_mem_dqinfo *info, uint blk, dqbuf_t buf) ++{ ++ struct super_block *sb = info->dqi_sb; ++ ++ return sb->s_op->quota_write(sb, info->dqi_type, (char *)buf, ++ info->dqi_usable_bs, blk << info->dqi_blocksize_bits); ++} ++ ++/* Remove empty block from list and return it */ ++static int get_free_dqblk(struct qtree_mem_dqinfo *info) ++{ ++ dqbuf_t buf = getdqbuf(info->dqi_usable_bs); ++ struct qt_disk_dqdbheader *dh = (struct qt_disk_dqdbheader *)buf; ++ int ret, blk; ++ ++ if (!buf) ++ return -ENOMEM; ++ if (info->dqi_free_blk) { ++ blk = info->dqi_free_blk; ++ ret = read_blk(info, blk, buf); ++ if (ret < 0) ++ goto out_buf; ++ info->dqi_free_blk = le32_to_cpu(dh->dqdh_next_free); ++ } ++ else { ++ memset(buf, 0, info->dqi_usable_bs); ++ /* Assure block allocation... */ ++ ret = write_blk(info, info->dqi_blocks, buf); ++ if (ret < 0) ++ goto out_buf; ++ blk = info->dqi_blocks++; ++ } ++ mark_info_dirty(info->dqi_sb, info->dqi_type); ++ ret = blk; ++out_buf: ++ freedqbuf(buf); ++ return ret; ++} ++ ++/* Insert empty block to the list */ ++static int put_free_dqblk(struct qtree_mem_dqinfo *info, dqbuf_t buf, uint blk) ++{ ++ struct qt_disk_dqdbheader *dh = (struct qt_disk_dqdbheader *)buf; ++ int err; ++ ++ dh->dqdh_next_free = cpu_to_le32(info->dqi_free_blk); ++ dh->dqdh_prev_free = cpu_to_le32(0); ++ dh->dqdh_entries = cpu_to_le16(0); ++ err = write_blk(info, blk, buf); ++ if (err < 0) ++ return err; ++ info->dqi_free_blk = blk; ++ mark_info_dirty(info->dqi_sb, info->dqi_type); ++ return 0; ++} ++ ++/* Remove given block from the list of blocks with free entries */ ++static int remove_free_dqentry(struct qtree_mem_dqinfo *info, dqbuf_t buf, uint blk) ++{ ++ dqbuf_t tmpbuf = getdqbuf(info->dqi_usable_bs); ++ struct qt_disk_dqdbheader *dh = (struct qt_disk_dqdbheader *)buf; ++ uint nextblk = le32_to_cpu(dh->dqdh_next_free); ++ uint prevblk = le32_to_cpu(dh->dqdh_prev_free); ++ int err; ++ ++ if (!tmpbuf) ++ return -ENOMEM; ++ if (nextblk) { ++ err = read_blk(info, nextblk, tmpbuf); ++ if (err < 0) ++ goto out_buf; ++ ((struct qt_disk_dqdbheader *)tmpbuf)->dqdh_prev_free = ++ dh->dqdh_prev_free; ++ err = write_blk(info, nextblk, tmpbuf); ++ if (err < 0) ++ goto out_buf; ++ } ++ if (prevblk) { ++ err = read_blk(info, prevblk, tmpbuf); ++ if (err < 0) ++ goto out_buf; ++ ((struct qt_disk_dqdbheader *)tmpbuf)->dqdh_next_free = ++ dh->dqdh_next_free; ++ err = write_blk(info, prevblk, tmpbuf); ++ if (err < 0) ++ goto out_buf; ++ } else { ++ info->dqi_free_entry = nextblk; ++ mark_info_dirty(info->dqi_sb, info->dqi_type); ++ } ++ freedqbuf(tmpbuf); ++ dh->dqdh_next_free = dh->dqdh_prev_free = cpu_to_le32(0); ++ /* No matter whether write succeeds block is out of list */ ++ if (write_blk(info, blk, buf) < 0) ++ printk(KERN_ERR "VFS: Can't write block (%u) with free entries.\n", blk); ++ return 0; ++out_buf: ++ freedqbuf(tmpbuf); ++ return err; ++} ++ ++/* Insert given block to the beginning of list with free entries */ ++static int insert_free_dqentry(struct qtree_mem_dqinfo *info, dqbuf_t buf, uint blk) ++{ ++ dqbuf_t tmpbuf = getdqbuf(info->dqi_usable_bs); ++ struct qt_disk_dqdbheader *dh = (struct qt_disk_dqdbheader *)buf; ++ int err; ++ ++ if (!tmpbuf) ++ return -ENOMEM; ++ dh->dqdh_next_free = cpu_to_le32(info->dqi_free_entry); ++ dh->dqdh_prev_free = cpu_to_le32(0); ++ err = write_blk(info, blk, buf); ++ if (err < 0) ++ goto out_buf; ++ if (info->dqi_free_entry) { ++ err = read_blk(info, info->dqi_free_entry, tmpbuf); ++ if (err < 0) ++ goto out_buf; ++ ((struct qt_disk_dqdbheader *)tmpbuf)->dqdh_prev_free = ++ cpu_to_le32(blk); ++ err = write_blk(info, info->dqi_free_entry, tmpbuf); ++ if (err < 0) ++ goto out_buf; ++ } ++ freedqbuf(tmpbuf); ++ info->dqi_free_entry = blk; ++ mark_info_dirty(info->dqi_sb, info->dqi_type); ++ return 0; ++out_buf: ++ freedqbuf(tmpbuf); ++ return err; ++} ++ ++/* Is the entry in the block free? */ ++int qtree_entry_unused(struct qtree_mem_dqinfo *info, char *disk) ++{ ++ int i; ++ ++ for (i = 0; i < info->dqi_entry_size; i++) ++ if (disk[i]) ++ return 0; ++ return 1; ++} ++EXPORT_SYMBOL(qtree_entry_unused); ++ ++/* Find space for dquot */ ++static uint find_free_dqentry(struct qtree_mem_dqinfo *info, ++ struct dquot *dquot, int *err) ++{ ++ uint blk, i; ++ struct qt_disk_dqdbheader *dh; ++ dqbuf_t buf = getdqbuf(info->dqi_usable_bs); ++ char *ddquot; ++ ++ *err = 0; ++ if (!buf) { ++ *err = -ENOMEM; ++ return 0; ++ } ++ dh = (struct qt_disk_dqdbheader *)buf; ++ if (info->dqi_free_entry) { ++ blk = info->dqi_free_entry; ++ *err = read_blk(info, blk, buf); ++ if (*err < 0) ++ goto out_buf; ++ } else { ++ blk = get_free_dqblk(info); ++ if ((int)blk < 0) { ++ *err = blk; ++ freedqbuf(buf); ++ return 0; ++ } ++ memset(buf, 0, info->dqi_usable_bs); ++ /* This is enough as block is already zeroed and entry list is empty... */ ++ info->dqi_free_entry = blk; ++ mark_info_dirty(dquot->dq_sb, dquot->dq_type); ++ } ++ /* Block will be full? */ ++ if (le16_to_cpu(dh->dqdh_entries) + 1 >= qtree_dqstr_in_blk(info)) { ++ *err = remove_free_dqentry(info, buf, blk); ++ if (*err < 0) { ++ printk(KERN_ERR "VFS: find_free_dqentry(): Can't " ++ "remove block (%u) from entry free list.\n", ++ blk); ++ goto out_buf; ++ } ++ } ++ le16_add_cpu(&dh->dqdh_entries, 1); ++ /* Find free structure in block */ ++ for (i = 0, ddquot = ((char *)buf) + sizeof(struct qt_disk_dqdbheader); ++ i < qtree_dqstr_in_blk(info) && !qtree_entry_unused(info, ddquot); ++ i++, ddquot += info->dqi_entry_size); ++#ifdef __QUOTA_QT_PARANOIA ++ if (i == qtree_dqstr_in_blk(info)) { ++ printk(KERN_ERR "VFS: find_free_dqentry(): Data block full " ++ "but it shouldn't.\n"); ++ *err = -EIO; ++ goto out_buf; ++ } ++#endif ++ *err = write_blk(info, blk, buf); ++ if (*err < 0) { ++ printk(KERN_ERR "VFS: find_free_dqentry(): Can't write quota " ++ "data block %u.\n", blk); ++ goto out_buf; ++ } ++ dquot->dq_off = (blk << info->dqi_blocksize_bits) + ++ sizeof(struct qt_disk_dqdbheader) + ++ i * info->dqi_entry_size; ++ freedqbuf(buf); ++ return blk; ++out_buf: ++ freedqbuf(buf); ++ return 0; ++} ++ ++/* Insert reference to structure into the trie */ ++static int do_insert_tree(struct qtree_mem_dqinfo *info, struct dquot *dquot, ++ uint *treeblk, int depth) ++{ ++ dqbuf_t buf = getdqbuf(info->dqi_usable_bs); ++ int ret = 0, newson = 0, newact = 0; ++ __le32 *ref; ++ uint newblk; ++ ++ if (!buf) ++ return -ENOMEM; ++ if (!*treeblk) { ++ ret = get_free_dqblk(info); ++ if (ret < 0) ++ goto out_buf; ++ *treeblk = ret; ++ memset(buf, 0, info->dqi_usable_bs); ++ newact = 1; ++ } else { ++ ret = read_blk(info, *treeblk, buf); ++ if (ret < 0) { ++ printk(KERN_ERR "VFS: Can't read tree quota block " ++ "%u.\n", *treeblk); ++ goto out_buf; ++ } ++ } ++ ref = (__le32 *)buf; ++ newblk = le32_to_cpu(ref[get_index(info, dquot->dq_id, depth)]); ++ if (!newblk) ++ newson = 1; ++ if (depth == info->dqi_qtree_depth - 1) { ++#ifdef __QUOTA_QT_PARANOIA ++ if (newblk) { ++ printk(KERN_ERR "VFS: Inserting already present quota " ++ "entry (block %u).\n", ++ le32_to_cpu(ref[get_index(info, ++ dquot->dq_id, depth)])); ++ ret = -EIO; ++ goto out_buf; ++ } ++#endif ++ newblk = find_free_dqentry(info, dquot, &ret); ++ } else { ++ ret = do_insert_tree(info, dquot, &newblk, depth+1); ++ } ++ if (newson && ret >= 0) { ++ ref[get_index(info, dquot->dq_id, depth)] = ++ cpu_to_le32(newblk); ++ ret = write_blk(info, *treeblk, buf); ++ } else if (newact && ret < 0) { ++ put_free_dqblk(info, buf, *treeblk); ++ } ++out_buf: ++ freedqbuf(buf); ++ return ret; ++} ++ ++/* Wrapper for inserting quota structure into tree */ ++static inline int dq_insert_tree(struct qtree_mem_dqinfo *info, ++ struct dquot *dquot) ++{ ++ int tmp = QT_TREEOFF; ++ return do_insert_tree(info, dquot, &tmp, 0); ++} ++ ++/* ++ * We don't have to be afraid of deadlocks as we never have quotas on quota files... ++ */ ++int qtree_write_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot) ++{ ++ int type = dquot->dq_type; ++ struct super_block *sb = dquot->dq_sb; ++ ssize_t ret; ++ dqbuf_t ddquot = getdqbuf(info->dqi_entry_size); ++ ++ if (!ddquot) ++ return -ENOMEM; ++ ++ /* dq_off is guarded by dqio_mutex */ ++ if (!dquot->dq_off) { ++ ret = dq_insert_tree(info, dquot); ++ if (ret < 0) { ++ printk(KERN_ERR "VFS: Error %zd occurred while " ++ "creating quota.\n", ret); ++ freedqbuf(ddquot); ++ return ret; ++ } ++ } ++ spin_lock(&dq_data_lock); ++ info->dqi_ops->mem2disk_dqblk(ddquot, dquot); ++ spin_unlock(&dq_data_lock); ++ ret = sb->s_op->quota_write(sb, type, (char *)ddquot, ++ info->dqi_entry_size, dquot->dq_off); ++ if (ret != info->dqi_entry_size) { ++ printk(KERN_WARNING "VFS: dquota write failed on dev %s\n", ++ sb->s_id); ++ if (ret >= 0) ++ ret = -ENOSPC; ++ } else { ++ ret = 0; ++ } ++ dqstats.writes++; ++ freedqbuf(ddquot); ++ ++ return ret; ++} ++EXPORT_SYMBOL(qtree_write_dquot); ++ ++/* Free dquot entry in data block */ ++static int free_dqentry(struct qtree_mem_dqinfo *info, struct dquot *dquot, ++ uint blk) ++{ ++ struct qt_disk_dqdbheader *dh; ++ dqbuf_t buf = getdqbuf(info->dqi_usable_bs); ++ int ret = 0; ++ ++ if (!buf) ++ return -ENOMEM; ++ if (dquot->dq_off >> info->dqi_blocksize_bits != blk) { ++ printk(KERN_ERR "VFS: Quota structure has offset to other " ++ "block (%u) than it should (%u).\n", blk, ++ (uint)(dquot->dq_off >> info->dqi_blocksize_bits)); ++ goto out_buf; ++ } ++ ret = read_blk(info, blk, buf); ++ if (ret < 0) { ++ printk(KERN_ERR "VFS: Can't read quota data block %u\n", blk); ++ goto out_buf; ++ } ++ dh = (struct qt_disk_dqdbheader *)buf; ++ le16_add_cpu(&dh->dqdh_entries, -1); ++ if (!le16_to_cpu(dh->dqdh_entries)) { /* Block got free? */ ++ ret = remove_free_dqentry(info, buf, blk); ++ if (ret >= 0) ++ ret = put_free_dqblk(info, buf, blk); ++ if (ret < 0) { ++ printk(KERN_ERR "VFS: Can't move quota data block (%u) " ++ "to free list.\n", blk); ++ goto out_buf; ++ } ++ } else { ++ memset(buf + ++ (dquot->dq_off & ((1 << info->dqi_blocksize_bits) - 1)), ++ 0, info->dqi_entry_size); ++ if (le16_to_cpu(dh->dqdh_entries) == ++ qtree_dqstr_in_blk(info) - 1) { ++ /* Insert will write block itself */ ++ ret = insert_free_dqentry(info, buf, blk); ++ if (ret < 0) { ++ printk(KERN_ERR "VFS: Can't insert quota data " ++ "block (%u) to free entry list.\n", blk); ++ goto out_buf; ++ } ++ } else { ++ ret = write_blk(info, blk, buf); ++ if (ret < 0) { ++ printk(KERN_ERR "VFS: Can't write quota data " ++ "block %u\n", blk); ++ goto out_buf; ++ } ++ } ++ } ++ dquot->dq_off = 0; /* Quota is now unattached */ ++out_buf: ++ freedqbuf(buf); ++ return ret; ++} ++ ++/* Remove reference to dquot from tree */ ++static int remove_tree(struct qtree_mem_dqinfo *info, struct dquot *dquot, ++ uint *blk, int depth) ++{ ++ dqbuf_t buf = getdqbuf(info->dqi_usable_bs); ++ int ret = 0; ++ uint newblk; ++ __le32 *ref = (__le32 *)buf; ++ ++ if (!buf) ++ return -ENOMEM; ++ ret = read_blk(info, *blk, buf); ++ if (ret < 0) { ++ printk(KERN_ERR "VFS: Can't read quota data block %u\n", *blk); ++ goto out_buf; ++ } ++ newblk = le32_to_cpu(ref[get_index(info, dquot->dq_id, depth)]); ++ if (depth == info->dqi_qtree_depth - 1) { ++ ret = free_dqentry(info, dquot, newblk); ++ newblk = 0; ++ } else { ++ ret = remove_tree(info, dquot, &newblk, depth+1); ++ } ++ if (ret >= 0 && !newblk) { ++ int i; ++ ref[get_index(info, dquot->dq_id, depth)] = cpu_to_le32(0); ++ /* Block got empty? */ ++ for (i = 0; ++ i < (info->dqi_usable_bs >> 2) && !ref[i]; ++ i++); ++ /* Don't put the root block into the free block list */ ++ if (i == (info->dqi_usable_bs >> 2) ++ && *blk != QT_TREEOFF) { ++ put_free_dqblk(info, buf, *blk); ++ *blk = 0; ++ } else { ++ ret = write_blk(info, *blk, buf); ++ if (ret < 0) ++ printk(KERN_ERR "VFS: Can't write quota tree " ++ "block %u.\n", *blk); ++ } ++ } ++out_buf: ++ freedqbuf(buf); ++ return ret; ++} ++ ++/* Delete dquot from tree */ ++int qtree_delete_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot) ++{ ++ uint tmp = QT_TREEOFF; ++ ++ if (!dquot->dq_off) /* Even not allocated? */ ++ return 0; ++ return remove_tree(info, dquot, &tmp, 0); ++} ++EXPORT_SYMBOL(qtree_delete_dquot); ++ ++/* Find entry in block */ ++static loff_t find_block_dqentry(struct qtree_mem_dqinfo *info, ++ struct dquot *dquot, uint blk) ++{ ++ dqbuf_t buf = getdqbuf(info->dqi_usable_bs); ++ loff_t ret = 0; ++ int i; ++ char *ddquot; ++ ++ if (!buf) ++ return -ENOMEM; ++ ret = read_blk(info, blk, buf); ++ if (ret < 0) { ++ printk(KERN_ERR "VFS: Can't read quota tree block %u.\n", blk); ++ goto out_buf; ++ } ++ for (i = 0, ddquot = ((char *)buf) + sizeof(struct qt_disk_dqdbheader); ++ i < qtree_dqstr_in_blk(info) && !info->dqi_ops->is_id(ddquot, dquot); ++ i++, ddquot += info->dqi_entry_size); ++ if (i == qtree_dqstr_in_blk(info)) { ++ printk(KERN_ERR "VFS: Quota for id %u referenced " ++ "but not present.\n", dquot->dq_id); ++ ret = -EIO; ++ goto out_buf; ++ } else { ++ ret = (blk << info->dqi_blocksize_bits) + sizeof(struct ++ qt_disk_dqdbheader) + i * info->dqi_entry_size; ++ } ++out_buf: ++ freedqbuf(buf); ++ return ret; ++} ++ ++/* Find entry for given id in the tree */ ++static loff_t find_tree_dqentry(struct qtree_mem_dqinfo *info, ++ struct dquot *dquot, uint blk, int depth) ++{ ++ dqbuf_t buf = getdqbuf(info->dqi_usable_bs); ++ loff_t ret = 0; ++ __le32 *ref = (__le32 *)buf; ++ ++ if (!buf) ++ return -ENOMEM; ++ ret = read_blk(info, blk, buf); ++ if (ret < 0) { ++ printk(KERN_ERR "VFS: Can't read quota tree block %u.\n", blk); ++ goto out_buf; ++ } ++ ret = 0; ++ blk = le32_to_cpu(ref[get_index(info, dquot->dq_id, depth)]); ++ if (!blk) /* No reference? */ ++ goto out_buf; ++ if (depth < info->dqi_qtree_depth - 1) ++ ret = find_tree_dqentry(info, dquot, blk, depth+1); ++ else ++ ret = find_block_dqentry(info, dquot, blk); ++out_buf: ++ freedqbuf(buf); ++ return ret; ++} ++ ++/* Find entry for given id in the tree - wrapper function */ ++static inline loff_t find_dqentry(struct qtree_mem_dqinfo *info, ++ struct dquot *dquot) ++{ ++ return find_tree_dqentry(info, dquot, QT_TREEOFF, 0); ++} ++ ++int qtree_read_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot) ++{ ++ int type = dquot->dq_type; ++ struct super_block *sb = dquot->dq_sb; ++ loff_t offset; ++ dqbuf_t ddquot; ++ int ret = 0; ++ ++#ifdef __QUOTA_QT_PARANOIA ++ /* Invalidated quota? */ ++ if (!sb_dqopt(dquot->dq_sb)->files[type]) { ++ printk(KERN_ERR "VFS: Quota invalidated while reading!\n"); ++ return -EIO; ++ } ++#endif ++ /* Do we know offset of the dquot entry in the quota file? */ ++ if (!dquot->dq_off) { ++ offset = find_dqentry(info, dquot); ++ if (offset <= 0) { /* Entry not present? */ ++ if (offset < 0) ++ printk(KERN_ERR "VFS: Can't read quota " ++ "structure for id %u.\n", dquot->dq_id); ++ dquot->dq_off = 0; ++ set_bit(DQ_FAKE_B, &dquot->dq_flags); ++ memset(&dquot->dq_dqb, 0, sizeof(struct mem_dqblk)); ++ ret = offset; ++ goto out; ++ } ++ dquot->dq_off = offset; ++ } ++ ddquot = getdqbuf(info->dqi_entry_size); ++ if (!ddquot) ++ return -ENOMEM; ++ ret = sb->s_op->quota_read(sb, type, (char *)ddquot, ++ info->dqi_entry_size, dquot->dq_off); ++ if (ret != info->dqi_entry_size) { ++ if (ret >= 0) ++ ret = -EIO; ++ printk(KERN_ERR "VFS: Error while reading quota " ++ "structure for id %u.\n", dquot->dq_id); ++ set_bit(DQ_FAKE_B, &dquot->dq_flags); ++ memset(&dquot->dq_dqb, 0, sizeof(struct mem_dqblk)); ++ freedqbuf(ddquot); ++ goto out; ++ } ++ spin_lock(&dq_data_lock); ++ info->dqi_ops->disk2mem_dqblk(dquot, ddquot); ++ if (!dquot->dq_dqb.dqb_bhardlimit && ++ !dquot->dq_dqb.dqb_bsoftlimit && ++ !dquot->dq_dqb.dqb_ihardlimit && ++ !dquot->dq_dqb.dqb_isoftlimit) ++ set_bit(DQ_FAKE_B, &dquot->dq_flags); ++ spin_unlock(&dq_data_lock); ++ freedqbuf(ddquot); ++out: ++ dqstats.reads++; ++ return ret; ++} ++EXPORT_SYMBOL(qtree_read_dquot); ++ ++/* Check whether dquot should not be deleted. We know we are ++ * the only one operating on dquot (thanks to dq_lock) */ ++int qtree_release_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot) ++{ ++ if (test_bit(DQ_FAKE_B, &dquot->dq_flags) && !(dquot->dq_dqb.dqb_curinodes | dquot->dq_dqb.dqb_curspace)) ++ return qtree_delete_dquot(info, dquot); ++ return 0; ++} ++EXPORT_SYMBOL(qtree_release_dquot); +Index: linux-2.6.27.36/fs/quota_tree.h +=================================================================== +--- /dev/null 1970-01-01 00:00:00.000000000 +0000 ++++ linux-2.6.27.36/fs/quota_tree.h 2009-10-08 16:32:48.000000000 +0200 +@@ -0,0 +1,25 @@ ++/* ++ * Definitions of structures for vfsv0 quota format ++ */ ++ ++#ifndef _LINUX_QUOTA_TREE_H ++#define _LINUX_QUOTA_TREE_H ++ ++#include ++#include ++ ++/* ++ * Structure of header of block with quota structures. It is padded to 16 bytes so ++ * there will be space for exactly 21 quota-entries in a block ++ */ ++struct qt_disk_dqdbheader { ++ __le32 dqdh_next_free; /* Number of next block with free entry */ ++ __le32 dqdh_prev_free; /* Number of previous block with free entry */ ++ __le16 dqdh_entries; /* Number of valid entries in block */ ++ __le16 dqdh_pad1; ++ __le32 dqdh_pad2; ++}; ++ ++#define QT_TREEOFF 1 /* Offset of tree in file in blocks */ ++ ++#endif /* _LINUX_QUOTAIO_TREE_H */ +Index: linux-2.6.27.36/fs/quota_v2.c +=================================================================== +--- linux-2.6.27.36.orig/fs/quota_v2.c 2009-10-05 17:19:01.000000000 +0200 ++++ linux-2.6.27.36/fs/quota_v2.c 2009-10-08 16:34:36.000000000 +0200 +@@ -6,6 +6,7 @@ + #include + #include + #include ++#include "quota_tree.h" + #include + #include + #include +@@ -15,16 +16,22 @@ + + #include + ++ + MODULE_AUTHOR("Jan Kara"); + MODULE_DESCRIPTION("Quota format v2 support"); + MODULE_LICENSE("GPL"); + + #define __QUOTA_V2_PARANOIA + +-typedef char *dqbuf_t; +- +-#define GETIDINDEX(id, depth) (((id) >> ((V2_DQTREEDEPTH-(depth)-1)*8)) & 0xff) +-#define GETENTRIES(buf) ((struct v2_disk_dqblk *)(((char *)buf)+sizeof(struct v2_disk_dqdbheader))) ++static void v2_mem2diskdqb(void *dp, struct dquot *dquot); ++static void v2_disk2memdqb(struct dquot *dquot, void *dp); ++static int v2_is_id(void *dp, struct dquot *dquot); ++ ++static struct qtree_fmt_operations v2_qtree_ops = { ++ .mem2disk_dqblk = v2_mem2diskdqb, ++ .disk2mem_dqblk = v2_disk2memdqb, ++ .is_id = v2_is_id, ++}; + + /* Check whether given file is really vfsv0 quotafile */ + static int v2_check_quota_file(struct super_block *sb, int type) +@@ -50,7 +57,7 @@ + static int v2_read_file_info(struct super_block *sb, int type) + { + struct v2_disk_dqinfo dinfo; +- struct mem_dqinfo *info = sb_dqopt(sb)->info+type; ++ struct mem_dqinfo *info = sb_dqinfo(sb, type); + ssize_t size; + + size = sb->s_op->quota_read(sb, type, (char *)&dinfo, +@@ -66,9 +73,16 @@ + info->dqi_bgrace = le32_to_cpu(dinfo.dqi_bgrace); + info->dqi_igrace = le32_to_cpu(dinfo.dqi_igrace); + info->dqi_flags = le32_to_cpu(dinfo.dqi_flags); +- info->u.v2_i.dqi_blocks = le32_to_cpu(dinfo.dqi_blocks); +- info->u.v2_i.dqi_free_blk = le32_to_cpu(dinfo.dqi_free_blk); +- info->u.v2_i.dqi_free_entry = le32_to_cpu(dinfo.dqi_free_entry); ++ info->u.v2_i.i.dqi_sb = sb; ++ info->u.v2_i.i.dqi_type = type; ++ info->u.v2_i.i.dqi_blocks = le32_to_cpu(dinfo.dqi_blocks); ++ info->u.v2_i.i.dqi_free_blk = le32_to_cpu(dinfo.dqi_free_blk); ++ info->u.v2_i.i.dqi_free_entry = le32_to_cpu(dinfo.dqi_free_entry); ++ info->u.v2_i.i.dqi_blocksize_bits = V2_DQBLKSIZE_BITS; ++ info->u.v2_i.i.dqi_usable_bs = 1 << V2_DQBLKSIZE_BITS; ++ info->u.v2_i.i.dqi_qtree_depth = qtree_depth(&info->u.v2_i.i); ++ info->u.v2_i.i.dqi_entry_size = sizeof(struct v2_disk_dqblk); ++ info->u.v2_i.i.dqi_ops = &v2_qtree_ops; + return 0; + } + +@@ -76,7 +90,7 @@ + static int v2_write_file_info(struct super_block *sb, int type) + { + struct v2_disk_dqinfo dinfo; +- struct mem_dqinfo *info = sb_dqopt(sb)->info+type; ++ struct mem_dqinfo *info = sb_dqinfo(sb, type); + ssize_t size; + + spin_lock(&dq_data_lock); +@@ -85,9 +99,9 @@ + dinfo.dqi_igrace = cpu_to_le32(info->dqi_igrace); + dinfo.dqi_flags = cpu_to_le32(info->dqi_flags & DQF_MASK); + spin_unlock(&dq_data_lock); +- dinfo.dqi_blocks = cpu_to_le32(info->u.v2_i.dqi_blocks); +- dinfo.dqi_free_blk = cpu_to_le32(info->u.v2_i.dqi_free_blk); +- dinfo.dqi_free_entry = cpu_to_le32(info->u.v2_i.dqi_free_entry); ++ dinfo.dqi_blocks = cpu_to_le32(info->u.v2_i.i.dqi_blocks); ++ dinfo.dqi_free_blk = cpu_to_le32(info->u.v2_i.i.dqi_free_blk); ++ dinfo.dqi_free_entry = cpu_to_le32(info->u.v2_i.i.dqi_free_entry); + size = sb->s_op->quota_write(sb, type, (char *)&dinfo, + sizeof(struct v2_disk_dqinfo), V2_DQINFOOFF); + if (size != sizeof(struct v2_disk_dqinfo)) { +@@ -98,8 +112,11 @@ + return 0; + } + +-static void disk2memdqb(struct mem_dqblk *m, struct v2_disk_dqblk *d) ++static void v2_disk2memdqb(struct dquot *dquot, void *dp) + { ++ struct v2_disk_dqblk *d = dp, empty; ++ struct mem_dqblk *m = &dquot->dq_dqb; ++ + m->dqb_ihardlimit = le32_to_cpu(d->dqb_ihardlimit); + m->dqb_isoftlimit = le32_to_cpu(d->dqb_isoftlimit); + m->dqb_curinodes = le32_to_cpu(d->dqb_curinodes); +@@ -108,10 +125,20 @@ + m->dqb_bsoftlimit = le32_to_cpu(d->dqb_bsoftlimit); + m->dqb_curspace = le64_to_cpu(d->dqb_curspace); + m->dqb_btime = le64_to_cpu(d->dqb_btime); ++ /* We need to escape back all-zero structure */ ++ memset(&empty, 0, sizeof(struct v2_disk_dqblk)); ++ empty.dqb_itime = cpu_to_le64(1); ++ if (!memcmp(&empty, dp, sizeof(struct v2_disk_dqblk))) ++ m->dqb_itime = 0; + } + +-static void mem2diskdqb(struct v2_disk_dqblk *d, struct mem_dqblk *m, qid_t id) ++static void v2_mem2diskdqb(void *dp, struct dquot *dquot) + { ++ struct v2_disk_dqblk *d = dp; ++ struct mem_dqblk *m = &dquot->dq_dqb; ++ struct qtree_mem_dqinfo *info = ++ &sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i; ++ + d->dqb_ihardlimit = cpu_to_le32(m->dqb_ihardlimit); + d->dqb_isoftlimit = cpu_to_le32(m->dqb_isoftlimit); + d->dqb_curinodes = cpu_to_le32(m->dqb_curinodes); +@@ -120,553 +147,35 @@ + d->dqb_bsoftlimit = cpu_to_le32(m->dqb_bsoftlimit); + d->dqb_curspace = cpu_to_le64(m->dqb_curspace); + d->dqb_btime = cpu_to_le64(m->dqb_btime); +- d->dqb_id = cpu_to_le32(id); +-} +- +-static dqbuf_t getdqbuf(void) +-{ +- dqbuf_t buf = kmalloc(V2_DQBLKSIZE, GFP_NOFS); +- if (!buf) +- printk(KERN_WARNING "VFS: Not enough memory for quota buffers.\n"); +- return buf; +-} +- +-static inline void freedqbuf(dqbuf_t buf) +-{ +- kfree(buf); +-} +- +-static inline ssize_t read_blk(struct super_block *sb, int type, uint blk, dqbuf_t buf) +-{ +- memset(buf, 0, V2_DQBLKSIZE); +- return sb->s_op->quota_read(sb, type, (char *)buf, +- V2_DQBLKSIZE, blk << V2_DQBLKSIZE_BITS); +-} +- +-static inline ssize_t write_blk(struct super_block *sb, int type, uint blk, dqbuf_t buf) +-{ +- return sb->s_op->quota_write(sb, type, (char *)buf, +- V2_DQBLKSIZE, blk << V2_DQBLKSIZE_BITS); ++ d->dqb_id = cpu_to_le32(dquot->dq_id); ++ if (qtree_entry_unused(info, dp)) ++ d->dqb_itime = cpu_to_le64(1); + } + +-/* Remove empty block from list and return it */ +-static int get_free_dqblk(struct super_block *sb, int type) ++static int v2_is_id(void *dp, struct dquot *dquot) + { +- dqbuf_t buf = getdqbuf(); +- struct mem_dqinfo *info = sb_dqinfo(sb, type); +- struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf; +- int ret, blk; ++ struct v2_disk_dqblk *d = dp; ++ struct qtree_mem_dqinfo *info = ++ &sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i; + +- if (!buf) +- return -ENOMEM; +- if (info->u.v2_i.dqi_free_blk) { +- blk = info->u.v2_i.dqi_free_blk; +- if ((ret = read_blk(sb, type, blk, buf)) < 0) +- goto out_buf; +- info->u.v2_i.dqi_free_blk = le32_to_cpu(dh->dqdh_next_free); +- } +- else { +- memset(buf, 0, V2_DQBLKSIZE); +- /* Assure block allocation... */ +- if ((ret = write_blk(sb, type, info->u.v2_i.dqi_blocks, buf)) < 0) +- goto out_buf; +- blk = info->u.v2_i.dqi_blocks++; +- } +- mark_info_dirty(sb, type); +- ret = blk; +-out_buf: +- freedqbuf(buf); +- return ret; +-} +- +-/* Insert empty block to the list */ +-static int put_free_dqblk(struct super_block *sb, int type, dqbuf_t buf, uint blk) +-{ +- struct mem_dqinfo *info = sb_dqinfo(sb, type); +- struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf; +- int err; +- +- dh->dqdh_next_free = cpu_to_le32(info->u.v2_i.dqi_free_blk); +- dh->dqdh_prev_free = cpu_to_le32(0); +- dh->dqdh_entries = cpu_to_le16(0); +- info->u.v2_i.dqi_free_blk = blk; +- mark_info_dirty(sb, type); +- /* Some strange block. We had better leave it... */ +- if ((err = write_blk(sb, type, blk, buf)) < 0) +- return err; +- return 0; +-} +- +-/* Remove given block from the list of blocks with free entries */ +-static int remove_free_dqentry(struct super_block *sb, int type, dqbuf_t buf, uint blk) +-{ +- dqbuf_t tmpbuf = getdqbuf(); +- struct mem_dqinfo *info = sb_dqinfo(sb, type); +- struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf; +- uint nextblk = le32_to_cpu(dh->dqdh_next_free), prevblk = le32_to_cpu(dh->dqdh_prev_free); +- int err; +- +- if (!tmpbuf) +- return -ENOMEM; +- if (nextblk) { +- if ((err = read_blk(sb, type, nextblk, tmpbuf)) < 0) +- goto out_buf; +- ((struct v2_disk_dqdbheader *)tmpbuf)->dqdh_prev_free = dh->dqdh_prev_free; +- if ((err = write_blk(sb, type, nextblk, tmpbuf)) < 0) +- goto out_buf; +- } +- if (prevblk) { +- if ((err = read_blk(sb, type, prevblk, tmpbuf)) < 0) +- goto out_buf; +- ((struct v2_disk_dqdbheader *)tmpbuf)->dqdh_next_free = dh->dqdh_next_free; +- if ((err = write_blk(sb, type, prevblk, tmpbuf)) < 0) +- goto out_buf; +- } +- else { +- info->u.v2_i.dqi_free_entry = nextblk; +- mark_info_dirty(sb, type); +- } +- freedqbuf(tmpbuf); +- dh->dqdh_next_free = dh->dqdh_prev_free = cpu_to_le32(0); +- /* No matter whether write succeeds block is out of list */ +- if (write_blk(sb, type, blk, buf) < 0) +- printk(KERN_ERR "VFS: Can't write block (%u) with free entries.\n", blk); +- return 0; +-out_buf: +- freedqbuf(tmpbuf); +- return err; +-} +- +-/* Insert given block to the beginning of list with free entries */ +-static int insert_free_dqentry(struct super_block *sb, int type, dqbuf_t buf, uint blk) +-{ +- dqbuf_t tmpbuf = getdqbuf(); +- struct mem_dqinfo *info = sb_dqinfo(sb, type); +- struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf; +- int err; +- +- if (!tmpbuf) +- return -ENOMEM; +- dh->dqdh_next_free = cpu_to_le32(info->u.v2_i.dqi_free_entry); +- dh->dqdh_prev_free = cpu_to_le32(0); +- if ((err = write_blk(sb, type, blk, buf)) < 0) +- goto out_buf; +- if (info->u.v2_i.dqi_free_entry) { +- if ((err = read_blk(sb, type, info->u.v2_i.dqi_free_entry, tmpbuf)) < 0) +- goto out_buf; +- ((struct v2_disk_dqdbheader *)tmpbuf)->dqdh_prev_free = cpu_to_le32(blk); +- if ((err = write_blk(sb, type, info->u.v2_i.dqi_free_entry, tmpbuf)) < 0) +- goto out_buf; +- } +- freedqbuf(tmpbuf); +- info->u.v2_i.dqi_free_entry = blk; +- mark_info_dirty(sb, type); +- return 0; +-out_buf: +- freedqbuf(tmpbuf); +- return err; +-} +- +-/* Find space for dquot */ +-static uint find_free_dqentry(struct dquot *dquot, int *err) +-{ +- struct super_block *sb = dquot->dq_sb; +- struct mem_dqinfo *info = sb_dqopt(sb)->info+dquot->dq_type; +- uint blk, i; +- struct v2_disk_dqdbheader *dh; +- struct v2_disk_dqblk *ddquot; +- struct v2_disk_dqblk fakedquot; +- dqbuf_t buf; +- +- *err = 0; +- if (!(buf = getdqbuf())) { +- *err = -ENOMEM; ++ if (qtree_entry_unused(info, dp)) + return 0; +- } +- dh = (struct v2_disk_dqdbheader *)buf; +- ddquot = GETENTRIES(buf); +- if (info->u.v2_i.dqi_free_entry) { +- blk = info->u.v2_i.dqi_free_entry; +- if ((*err = read_blk(sb, dquot->dq_type, blk, buf)) < 0) +- goto out_buf; +- } +- else { +- blk = get_free_dqblk(sb, dquot->dq_type); +- if ((int)blk < 0) { +- *err = blk; +- freedqbuf(buf); +- return 0; +- } +- memset(buf, 0, V2_DQBLKSIZE); +- /* This is enough as block is already zeroed and entry list is empty... */ +- info->u.v2_i.dqi_free_entry = blk; +- mark_info_dirty(sb, dquot->dq_type); +- } +- if (le16_to_cpu(dh->dqdh_entries)+1 >= V2_DQSTRINBLK) /* Block will be full? */ +- if ((*err = remove_free_dqentry(sb, dquot->dq_type, buf, blk)) < 0) { +- printk(KERN_ERR "VFS: find_free_dqentry(): Can't remove block (%u) from entry free list.\n", blk); +- goto out_buf; +- } +- le16_add_cpu(&dh->dqdh_entries, 1); +- memset(&fakedquot, 0, sizeof(struct v2_disk_dqblk)); +- /* Find free structure in block */ +- for (i = 0; i < V2_DQSTRINBLK && memcmp(&fakedquot, ddquot+i, sizeof(struct v2_disk_dqblk)); i++); +-#ifdef __QUOTA_V2_PARANOIA +- if (i == V2_DQSTRINBLK) { +- printk(KERN_ERR "VFS: find_free_dqentry(): Data block full but it shouldn't.\n"); +- *err = -EIO; +- goto out_buf; +- } +-#endif +- if ((*err = write_blk(sb, dquot->dq_type, blk, buf)) < 0) { +- printk(KERN_ERR "VFS: find_free_dqentry(): Can't write quota data block %u.\n", blk); +- goto out_buf; +- } +- dquot->dq_off = (blk<dq_sb; +- dqbuf_t buf; +- int ret = 0, newson = 0, newact = 0; +- __le32 *ref; +- uint newblk; +- +- if (!(buf = getdqbuf())) +- return -ENOMEM; +- if (!*treeblk) { +- ret = get_free_dqblk(sb, dquot->dq_type); +- if (ret < 0) +- goto out_buf; +- *treeblk = ret; +- memset(buf, 0, V2_DQBLKSIZE); +- newact = 1; +- } +- else { +- if ((ret = read_blk(sb, dquot->dq_type, *treeblk, buf)) < 0) { +- printk(KERN_ERR "VFS: Can't read tree quota block %u.\n", *treeblk); +- goto out_buf; +- } +- } +- ref = (__le32 *)buf; +- newblk = le32_to_cpu(ref[GETIDINDEX(dquot->dq_id, depth)]); +- if (!newblk) +- newson = 1; +- if (depth == V2_DQTREEDEPTH-1) { +-#ifdef __QUOTA_V2_PARANOIA +- if (newblk) { +- printk(KERN_ERR "VFS: Inserting already present quota entry (block %u).\n", le32_to_cpu(ref[GETIDINDEX(dquot->dq_id, depth)])); +- ret = -EIO; +- goto out_buf; +- } +-#endif +- newblk = find_free_dqentry(dquot, &ret); +- } +- else +- ret = do_insert_tree(dquot, &newblk, depth+1); +- if (newson && ret >= 0) { +- ref[GETIDINDEX(dquot->dq_id, depth)] = cpu_to_le32(newblk); +- ret = write_blk(sb, dquot->dq_type, *treeblk, buf); +- } +- else if (newact && ret < 0) +- put_free_dqblk(sb, dquot->dq_type, buf, *treeblk); +-out_buf: +- freedqbuf(buf); +- return ret; ++ return le32_to_cpu(d->dqb_id) == dquot->dq_id; + } + +-/* Wrapper for inserting quota structure into tree */ +-static inline int dq_insert_tree(struct dquot *dquot) ++static int v2_read_dquot(struct dquot *dquot) + { +- int tmp = V2_DQTREEOFF; +- return do_insert_tree(dquot, &tmp, 0); ++ return qtree_read_dquot(&sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i, dquot); + } + +-/* +- * We don't have to be afraid of deadlocks as we never have quotas on quota files... +- */ + static int v2_write_dquot(struct dquot *dquot) + { +- int type = dquot->dq_type; +- ssize_t ret; +- struct v2_disk_dqblk ddquot, empty; +- +- /* dq_off is guarded by dqio_mutex */ +- if (!dquot->dq_off) +- if ((ret = dq_insert_tree(dquot)) < 0) { +- printk(KERN_ERR "VFS: Error %zd occurred while creating quota.\n", ret); +- return ret; +- } +- spin_lock(&dq_data_lock); +- mem2diskdqb(&ddquot, &dquot->dq_dqb, dquot->dq_id); +- /* Argh... We may need to write structure full of zeroes but that would be +- * treated as an empty place by the rest of the code. Format change would +- * be definitely cleaner but the problems probably are not worth it */ +- memset(&empty, 0, sizeof(struct v2_disk_dqblk)); +- if (!memcmp(&empty, &ddquot, sizeof(struct v2_disk_dqblk))) +- ddquot.dqb_itime = cpu_to_le64(1); +- spin_unlock(&dq_data_lock); +- ret = dquot->dq_sb->s_op->quota_write(dquot->dq_sb, type, +- (char *)&ddquot, sizeof(struct v2_disk_dqblk), dquot->dq_off); +- if (ret != sizeof(struct v2_disk_dqblk)) { +- printk(KERN_WARNING "VFS: dquota write failed on dev %s\n", dquot->dq_sb->s_id); +- if (ret >= 0) +- ret = -ENOSPC; +- } +- else +- ret = 0; +- dqstats.writes++; +- +- return ret; +-} +- +-/* Free dquot entry in data block */ +-static int free_dqentry(struct dquot *dquot, uint blk) +-{ +- struct super_block *sb = dquot->dq_sb; +- int type = dquot->dq_type; +- struct v2_disk_dqdbheader *dh; +- dqbuf_t buf = getdqbuf(); +- int ret = 0; +- +- if (!buf) +- return -ENOMEM; +- if (dquot->dq_off >> V2_DQBLKSIZE_BITS != blk) { +- printk(KERN_ERR "VFS: Quota structure has offset to other " +- "block (%u) than it should (%u).\n", blk, +- (uint)(dquot->dq_off >> V2_DQBLKSIZE_BITS)); +- goto out_buf; +- } +- if ((ret = read_blk(sb, type, blk, buf)) < 0) { +- printk(KERN_ERR "VFS: Can't read quota data block %u\n", blk); +- goto out_buf; +- } +- dh = (struct v2_disk_dqdbheader *)buf; +- le16_add_cpu(&dh->dqdh_entries, -1); +- if (!le16_to_cpu(dh->dqdh_entries)) { /* Block got free? */ +- if ((ret = remove_free_dqentry(sb, type, buf, blk)) < 0 || +- (ret = put_free_dqblk(sb, type, buf, blk)) < 0) { +- printk(KERN_ERR "VFS: Can't move quota data block (%u) " +- "to free list.\n", blk); +- goto out_buf; +- } +- } +- else { +- memset(buf+(dquot->dq_off & ((1 << V2_DQBLKSIZE_BITS)-1)), 0, +- sizeof(struct v2_disk_dqblk)); +- if (le16_to_cpu(dh->dqdh_entries) == V2_DQSTRINBLK-1) { +- /* Insert will write block itself */ +- if ((ret = insert_free_dqentry(sb, type, buf, blk)) < 0) { +- printk(KERN_ERR "VFS: Can't insert quota data block (%u) to free entry list.\n", blk); +- goto out_buf; +- } +- } +- else +- if ((ret = write_blk(sb, type, blk, buf)) < 0) { +- printk(KERN_ERR "VFS: Can't write quota data " +- "block %u\n", blk); +- goto out_buf; +- } +- } +- dquot->dq_off = 0; /* Quota is now unattached */ +-out_buf: +- freedqbuf(buf); +- return ret; +-} +- +-/* Remove reference to dquot from tree */ +-static int remove_tree(struct dquot *dquot, uint *blk, int depth) +-{ +- struct super_block *sb = dquot->dq_sb; +- int type = dquot->dq_type; +- dqbuf_t buf = getdqbuf(); +- int ret = 0; +- uint newblk; +- __le32 *ref = (__le32 *)buf; +- +- if (!buf) +- return -ENOMEM; +- if ((ret = read_blk(sb, type, *blk, buf)) < 0) { +- printk(KERN_ERR "VFS: Can't read quota data block %u\n", *blk); +- goto out_buf; +- } +- newblk = le32_to_cpu(ref[GETIDINDEX(dquot->dq_id, depth)]); +- if (depth == V2_DQTREEDEPTH-1) { +- ret = free_dqentry(dquot, newblk); +- newblk = 0; +- } +- else +- ret = remove_tree(dquot, &newblk, depth+1); +- if (ret >= 0 && !newblk) { +- int i; +- ref[GETIDINDEX(dquot->dq_id, depth)] = cpu_to_le32(0); +- for (i = 0; i < V2_DQBLKSIZE && !buf[i]; i++); /* Block got empty? */ +- /* Don't put the root block into the free block list */ +- if (i == V2_DQBLKSIZE && *blk != V2_DQTREEOFF) { +- put_free_dqblk(sb, type, buf, *blk); +- *blk = 0; +- } +- else +- if ((ret = write_blk(sb, type, *blk, buf)) < 0) +- printk(KERN_ERR "VFS: Can't write quota tree " +- "block %u.\n", *blk); +- } +-out_buf: +- freedqbuf(buf); +- return ret; +-} +- +-/* Delete dquot from tree */ +-static int v2_delete_dquot(struct dquot *dquot) +-{ +- uint tmp = V2_DQTREEOFF; +- +- if (!dquot->dq_off) /* Even not allocated? */ +- return 0; +- return remove_tree(dquot, &tmp, 0); +-} +- +-/* Find entry in block */ +-static loff_t find_block_dqentry(struct dquot *dquot, uint blk) +-{ +- dqbuf_t buf = getdqbuf(); +- loff_t ret = 0; +- int i; +- struct v2_disk_dqblk *ddquot = GETENTRIES(buf); +- +- if (!buf) +- return -ENOMEM; +- if ((ret = read_blk(dquot->dq_sb, dquot->dq_type, blk, buf)) < 0) { +- printk(KERN_ERR "VFS: Can't read quota tree block %u.\n", blk); +- goto out_buf; +- } +- if (dquot->dq_id) +- for (i = 0; i < V2_DQSTRINBLK && +- le32_to_cpu(ddquot[i].dqb_id) != dquot->dq_id; i++); +- else { /* ID 0 as a bit more complicated searching... */ +- struct v2_disk_dqblk fakedquot; +- +- memset(&fakedquot, 0, sizeof(struct v2_disk_dqblk)); +- for (i = 0; i < V2_DQSTRINBLK; i++) +- if (!le32_to_cpu(ddquot[i].dqb_id) && +- memcmp(&fakedquot, ddquot+i, sizeof(struct v2_disk_dqblk))) +- break; +- } +- if (i == V2_DQSTRINBLK) { +- printk(KERN_ERR "VFS: Quota for id %u referenced " +- "but not present.\n", dquot->dq_id); +- ret = -EIO; +- goto out_buf; +- } +- else +- ret = (blk << V2_DQBLKSIZE_BITS) + sizeof(struct +- v2_disk_dqdbheader) + i * sizeof(struct v2_disk_dqblk); +-out_buf: +- freedqbuf(buf); +- return ret; +-} +- +-/* Find entry for given id in the tree */ +-static loff_t find_tree_dqentry(struct dquot *dquot, uint blk, int depth) +-{ +- dqbuf_t buf = getdqbuf(); +- loff_t ret = 0; +- __le32 *ref = (__le32 *)buf; +- +- if (!buf) +- return -ENOMEM; +- if ((ret = read_blk(dquot->dq_sb, dquot->dq_type, blk, buf)) < 0) { +- printk(KERN_ERR "VFS: Can't read quota tree block %u.\n", blk); +- goto out_buf; +- } +- ret = 0; +- blk = le32_to_cpu(ref[GETIDINDEX(dquot->dq_id, depth)]); +- if (!blk) /* No reference? */ +- goto out_buf; +- if (depth < V2_DQTREEDEPTH-1) +- ret = find_tree_dqentry(dquot, blk, depth+1); +- else +- ret = find_block_dqentry(dquot, blk); +-out_buf: +- freedqbuf(buf); +- return ret; ++ return qtree_write_dquot(&sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i, dquot); + } + +-/* Find entry for given id in the tree - wrapper function */ +-static inline loff_t find_dqentry(struct dquot *dquot) +-{ +- return find_tree_dqentry(dquot, V2_DQTREEOFF, 0); +-} +- +-static int v2_read_dquot(struct dquot *dquot) +-{ +- int type = dquot->dq_type; +- loff_t offset; +- struct v2_disk_dqblk ddquot, empty; +- int ret = 0; +- +-#ifdef __QUOTA_V2_PARANOIA +- /* Invalidated quota? */ +- if (!dquot->dq_sb || !sb_dqopt(dquot->dq_sb)->files[type]) { +- printk(KERN_ERR "VFS: Quota invalidated while reading!\n"); +- return -EIO; +- } +-#endif +- offset = find_dqentry(dquot); +- if (offset <= 0) { /* Entry not present? */ +- if (offset < 0) +- printk(KERN_ERR "VFS: Can't read quota " +- "structure for id %u.\n", dquot->dq_id); +- dquot->dq_off = 0; +- set_bit(DQ_FAKE_B, &dquot->dq_flags); +- memset(&dquot->dq_dqb, 0, sizeof(struct mem_dqblk)); +- ret = offset; +- } +- else { +- dquot->dq_off = offset; +- if ((ret = dquot->dq_sb->s_op->quota_read(dquot->dq_sb, type, +- (char *)&ddquot, sizeof(struct v2_disk_dqblk), offset)) +- != sizeof(struct v2_disk_dqblk)) { +- if (ret >= 0) +- ret = -EIO; +- printk(KERN_ERR "VFS: Error while reading quota " +- "structure for id %u.\n", dquot->dq_id); +- memset(&ddquot, 0, sizeof(struct v2_disk_dqblk)); +- } +- else { +- ret = 0; +- /* We need to escape back all-zero structure */ +- memset(&empty, 0, sizeof(struct v2_disk_dqblk)); +- empty.dqb_itime = cpu_to_le64(1); +- if (!memcmp(&empty, &ddquot, sizeof(struct v2_disk_dqblk))) +- ddquot.dqb_itime = 0; +- } +- disk2memdqb(&dquot->dq_dqb, &ddquot); +- if (!dquot->dq_dqb.dqb_bhardlimit && +- !dquot->dq_dqb.dqb_bsoftlimit && +- !dquot->dq_dqb.dqb_ihardlimit && +- !dquot->dq_dqb.dqb_isoftlimit) +- set_bit(DQ_FAKE_B, &dquot->dq_flags); +- } +- dqstats.reads++; +- +- return ret; +-} +- +-/* Check whether dquot should not be deleted. We know we are +- * the only one operating on dquot (thanks to dq_lock) */ + static int v2_release_dquot(struct dquot *dquot) + { +- if (test_bit(DQ_FAKE_B, &dquot->dq_flags) && !(dquot->dq_dqb.dqb_curinodes | dquot->dq_dqb.dqb_curspace)) +- return v2_delete_dquot(dquot); +- return 0; ++ return qtree_release_dquot(&sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i, dquot); + } + + static struct quota_format_ops v2_format_ops = { +Index: linux-2.6.27.36/include/linux/dqblk_qtree.h +=================================================================== +--- /dev/null 1970-01-01 00:00:00.000000000 +0000 ++++ linux-2.6.27.36/include/linux/dqblk_qtree.h 2009-10-08 16:32:48.000000000 +0200 +@@ -0,0 +1,56 @@ ++/* ++ * Definitions of structures and functions for quota formats using trie ++ */ ++ ++#ifndef _LINUX_DQBLK_QTREE_H ++#define _LINUX_DQBLK_QTREE_H ++ ++#include ++ ++/* Numbers of blocks needed for updates - we count with the smallest ++ * possible block size (1024) */ ++#define QTREE_INIT_ALLOC 4 ++#define QTREE_INIT_REWRITE 2 ++#define QTREE_DEL_ALLOC 0 ++#define QTREE_DEL_REWRITE 6 ++ ++struct dquot; ++ ++/* Operations */ ++struct qtree_fmt_operations { ++ void (*mem2disk_dqblk)(void *disk, struct dquot *dquot); /* Convert given entry from in memory format to disk one */ ++ void (*disk2mem_dqblk)(struct dquot *dquot, void *disk); /* Convert given entry from disk format to in memory one */ ++ int (*is_id)(void *disk, struct dquot *dquot); /* Is this structure for given id? */ ++}; ++ ++/* Inmemory copy of version specific information */ ++struct qtree_mem_dqinfo { ++ struct super_block *dqi_sb; /* Sb quota is on */ ++ int dqi_type; /* Quota type */ ++ unsigned int dqi_blocks; /* # of blocks in quota file */ ++ unsigned int dqi_free_blk; /* First block in list of free blocks */ ++ unsigned int dqi_free_entry; /* First block with free entry */ ++ unsigned int dqi_blocksize_bits; /* Block size of quota file */ ++ unsigned int dqi_entry_size; /* Size of quota entry in quota file */ ++ unsigned int dqi_usable_bs; /* Space usable in block for quota data */ ++ unsigned int dqi_qtree_depth; /* Precomputed depth of quota tree */ ++ struct qtree_fmt_operations *dqi_ops; /* Operations for entry manipulation */ ++}; ++ ++int qtree_write_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot); ++int qtree_read_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot); ++int qtree_delete_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot); ++int qtree_release_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot); ++int qtree_entry_unused(struct qtree_mem_dqinfo *info, char *disk); ++static inline int qtree_depth(struct qtree_mem_dqinfo *info) ++{ ++ unsigned int epb = info->dqi_usable_bs >> 2; ++ unsigned long long entries = epb; ++ int i; ++ ++ for (i = 1; entries < (1ULL << 32); i++) ++ entries *= epb; ++ return i; ++} ++ ++#endif /* _LINUX_DQBLK_QTREE_H */ +Index: linux-2.6.27.36/include/linux/dqblk_v2.h +=================================================================== +--- linux-2.6.27.36.orig/include/linux/dqblk_v2.h 2009-10-05 17:19:01.000000000 +0200 ++++ linux-2.6.27.36/include/linux/dqblk_v2.h 2009-10-08 16:32:48.000000000 +0200 +@@ -1,26 +1,23 @@ + /* +- * Definitions of structures for vfsv0 quota format ++ * Definitions for vfsv0 quota format + */ + + #ifndef _LINUX_DQBLK_V2_H + #define _LINUX_DQBLK_V2_H + +-#include ++#include + +-/* id numbers of quota format */ ++/* Id number of quota format */ + #define QFMT_VFS_V0 2 + + /* Numbers of blocks needed for updates */ +-#define V2_INIT_ALLOC 4 +-#define V2_INIT_REWRITE 2 +-#define V2_DEL_ALLOC 0 +-#define V2_DEL_REWRITE 6 ++#define V2_INIT_ALLOC QTREE_INIT_ALLOC ++#define V2_INIT_REWRITE QTREE_INIT_REWRITE ++#define V2_DEL_ALLOC QTREE_DEL_ALLOC ++#define V2_DEL_REWRITE QTREE_DEL_REWRITE + +-/* Inmemory copy of version specific information */ + struct v2_mem_dqinfo { +- unsigned int dqi_blocks; +- unsigned int dqi_free_blk; +- unsigned int dqi_free_entry; ++ struct qtree_mem_dqinfo i; + }; + + #endif /* _LINUX_DQBLK_V2_H */ --- lustre-1.8.3.orig/lustre/kernel_patches/patches/sd_iostats-2.6.26-vanilla.patch +++ lustre-1.8.3/lustre/kernel_patches/patches/sd_iostats-2.6.26-vanilla.patch @@ -0,0 +1,579 @@ +Index: linux-source-2.6.26/drivers/scsi/Kconfig +=================================================================== +--- linux-source-2.6.26.orig/drivers/scsi/Kconfig 2009-10-09 13:33:25.000000000 +0200 ++++ linux-source-2.6.26/drivers/scsi/Kconfig 2009-10-09 13:39:06.000000000 +0200 +@@ -81,6 +81,14 @@ + In this case, do not compile the driver for your SCSI host adapter + (below) as a module either. + ++config SD_IOSTATS ++ bool "Enable SCSI disk I/O stats" ++ depends on BLK_DEV_SD ++ default y ++ ---help--- ++ This enables SCSI disk I/O stats collection. You must also enable ++ /proc file system support if you want this feature. ++ + config CHR_DEV_ST + tristate "SCSI tape support" + depends on SCSI +Index: linux-source-2.6.26/drivers/scsi/scsi_proc.c +=================================================================== +--- linux-source-2.6.26.orig/drivers/scsi/scsi_proc.c 2009-10-09 13:33:25.000000000 +0200 ++++ linux-source-2.6.26/drivers/scsi/scsi_proc.c 2009-10-09 13:39:06.000000000 +0200 +@@ -40,7 +40,8 @@ + /* 4K page size, but our output routines, use some slack for overruns */ + #define PROC_BLOCK_SIZE (3*1024) + +-static struct proc_dir_entry *proc_scsi; ++struct proc_dir_entry *proc_scsi; ++EXPORT_SYMBOL(proc_scsi); + + /* Protect sht->present and sht->proc_dir */ + static DEFINE_MUTEX(global_host_template_mutex); +Index: linux-source-2.6.26/drivers/scsi/sd.c +=================================================================== +--- linux-source-2.6.26.orig/drivers/scsi/sd.c 2009-10-09 13:33:25.000000000 +0200 ++++ linux-source-2.6.26/drivers/scsi/sd.c 2009-10-09 13:39:50.000000000 +0200 +@@ -107,6 +107,24 @@ + * object after last put) */ + static DEFINE_MUTEX(sd_ref_mutex); + ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++# include ++# include ++struct proc_dir_entry *sd_iostats_procdir = NULL; ++char sd_iostats_procdir_name[] = "sd_iostats"; ++static struct file_operations sd_iostats_proc_fops; ++ ++extern void sd_iostats_init(void); ++extern void sd_iostats_fini(void); ++void sd_iostats_start_req(struct scsi_cmnd *SCpnt); ++void sd_iostats_finish_req(struct scsi_cmnd *SCpnt); ++#else ++static inline void sd_iostats_init(void) {} ++static inline void sd_iostats_fini(void) {} ++static inline void sd_iostats_start_req(struct scsi_cmnd *SCpnt) {} ++static inline void sd_iostats_finish_req(struct scsi_cmnd *SCpnt) {} ++#endif ++ + static const char *sd_cache_types[] = { + "write through", "none", "write back", + "write back, no read (daft)" +@@ -578,6 +596,8 @@ + + sdev = sdkp->device; + ++ sd_iostats_start_req(SCpnt); ++ + /* + * If the device is in error recovery, wait until it is done. + * If the device is offline, then disallow any access to it. +@@ -1023,6 +1043,7 @@ + break; + } + out: ++ sd_iostats_finish_req(SCpnt); + return good_bytes; + } + +@@ -1712,6 +1733,36 @@ + if (sdp->removable) + gd->flags |= GENHD_FL_REMOVABLE; + ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++ sdkp->stats = kzalloc(sizeof(iostat_stats_t), GFP_KERNEL); ++ if (!sdkp->stats) { ++ printk(KERN_WARNING "cannot allocate iostat structure for" ++ "%s\n", gd->disk_name); ++ } else { ++ do_gettimeofday(&sdkp->stats->iostat_timeval); ++ sdkp->stats->iostat_queue_stamp = jiffies; ++ spin_lock_init(&sdkp->stats->iostat_lock); ++ if (sd_iostats_procdir) { ++ struct proc_dir_entry *pde; ++ pde = create_proc_entry(gd->disk_name, S_IRUGO | S_IWUSR, ++ sd_iostats_procdir); ++ if (!pde) { ++ printk(KERN_WARNING "Can't create /proc/scsi/" ++ "%s/%s\n", ++ sd_iostats_procdir_name, ++ gd->disk_name); ++ kfree(sdkp->stats); ++ sdkp->stats = NULL; ++ } else { ++ pde->proc_fops = &sd_iostats_proc_fops; ++ pde->data = gd; ++ } ++ } else { ++ kfree(sdkp->stats); ++ sdkp->stats = NULL; ++ } ++ } ++#endif + dev_set_drvdata(dev, sdkp); + add_disk(gd); + +@@ -1755,6 +1806,366 @@ + return 0; + } + ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++static int ++sd_iostats_seq_show(struct seq_file *seq, void *v) ++{ ++ struct timeval now; ++ struct gendisk *disk = seq->private; ++ iostat_stats_t *stats; ++ unsigned long long read_len; ++ unsigned long long read_len_tot; ++ unsigned long read_num; ++ unsigned long read_num_tot; ++ unsigned long long write_len; ++ unsigned long long write_len_tot; ++ unsigned long write_num; ++ unsigned long write_num_tot; ++ int i; ++ int maxi; ++ ++ stats = scsi_disk(disk)->stats; ++ if (stats == NULL) { ++ printk(KERN_ERR "sd_iostats_seq_show: NULL stats entry\n"); ++ BUG(); ++ } ++ ++ do_gettimeofday(&now); ++ now.tv_sec -= stats->iostat_timeval.tv_sec; ++ now.tv_usec -= stats->iostat_timeval.tv_usec; ++ if (now.tv_usec < 0) { ++ now.tv_usec += 1000000; ++ now.tv_sec--; ++ } ++ ++ /* this sampling races with updates */ ++ seq_printf(seq, "index: %lu snapshot_time: %lu.%06lu\n", ++ (unsigned long) scsi_disk(disk)->index, ++ now.tv_sec, now.tv_usec); ++ ++ for (i = IOSTAT_NCOUNTERS - 1; i > 0; i--) ++ if (stats->iostat_read_histogram[i].iostat_count != 0 || ++ stats->iostat_write_histogram[i].iostat_count != 0) ++ break; ++ maxi = i; ++ ++ seq_printf(seq, "%8s %8s %12s %8s %12s\n", "size", ++ "reads", "total", "writes", "total"); ++ ++ read_len_tot = write_len_tot = 0; ++ read_num_tot = write_num_tot = 0; ++ for (i = 0; i <= maxi; i++) { ++ read_len = stats->iostat_read_histogram[i].iostat_size; ++ read_len_tot += read_len; ++ read_num = stats->iostat_read_histogram[i].iostat_count; ++ read_num_tot += read_num; ++ ++ write_len = stats->iostat_write_histogram[i].iostat_size; ++ write_len_tot += write_len; ++ write_num = stats->iostat_write_histogram[i].iostat_count; ++ write_num_tot += write_num; ++ ++ seq_printf (seq, "%8d %8lu %12llu %8lu %12llu\n", ++ 512<iostat_queue_ticks[i]; ++ if (ticks == 0) ++ continue; ++ percent = stats->iostat_queue_ticks[i] * 100; ++ do_div(percent, stats->iostat_queue_ticks_sum); ++ seq_printf(seq, "%8d %8llu %8llu\n", i, ticks, percent); ++ } ++ ++ if (stats->iostat_reqs != 0) { ++ unsigned long long aveseek = 0, percent = 0; ++ ++ if (stats->iostat_seeks) { ++ aveseek = stats->iostat_seek_sectors; ++ do_div(aveseek, stats->iostat_seeks); ++ percent = stats->iostat_seeks * 100; ++ do_div(percent, stats->iostat_reqs); ++ } ++ ++ seq_printf(seq, "\n%llu sectors in %llu reqs: %llu seek(s) over " ++ "%llu sectors in ave, %llu%% of all reqs\n", ++ stats->iostat_sectors, stats->iostat_reqs, ++ stats->iostat_seeks, aveseek, percent); ++ } ++ ++ seq_printf(seq, "\n%16s %8s %8s %8s %8s\n", "process time", "reads", ++ "%%", "writes", "%%"); ++ for (i = 0; i < IOSTAT_NCOUNTERS; i++) { ++ unsigned long read_percent = 0, write_percent = 0; ++ if (stats->iostat_wtime[i] == 0 && ++ stats->iostat_rtime[i] == 0) ++ continue; ++ if (stats->iostat_read_reqs) ++ read_percent = stats->iostat_rtime[i] * 100 / ++ stats->iostat_read_reqs; ++ if (stats->iostat_write_reqs) ++ write_percent = stats->iostat_wtime[i] * 100 / ++ stats->iostat_write_reqs; ++ seq_printf(seq, "%16u %8lu %8lu %8lu %8lu\n", ++ jiffies_to_msecs(((1UL << i) >> 1) << 1), ++ stats->iostat_rtime[i], read_percent, ++ stats->iostat_wtime[i], write_percent); ++ } ++ ++ seq_printf(seq, "\n%16s %8s %8s %8s %8s\n", "time in queue", "reads", ++ "%%", "writes", "%%"); ++ for (i = 0; i < IOSTAT_NCOUNTERS; i++) { ++ unsigned long read_percent = 0, write_percent = 0; ++ if (stats->iostat_wtime_in_queue[i] == 0 && ++ stats->iostat_rtime_in_queue[i] == 0) ++ continue; ++ if (stats->iostat_read_reqs) ++ read_percent = stats->iostat_rtime_in_queue[i] * 100 / ++ stats->iostat_read_reqs; ++ if (stats->iostat_write_reqs) ++ write_percent = stats->iostat_wtime_in_queue[i] * 100 / ++ stats->iostat_write_reqs; ++ seq_printf(seq, "%16u %8lu %8lu %8lu %8lu\n", ++ jiffies_to_msecs(((1UL << i) >> 1) << 1), ++ stats->iostat_rtime_in_queue[i], ++ read_percent, ++ stats->iostat_wtime_in_queue[i], ++ write_percent); ++ } ++ ++ return 0; ++} ++ ++static void * ++sd_iostats_seq_start(struct seq_file *p, loff_t *pos) ++{ ++ return (*pos == 0) ? (void *)1 : NULL; ++} ++ ++static void * ++sd_iostats_seq_next(struct seq_file *p, void *v, loff_t *pos) ++{ ++ ++*pos; ++ return NULL; ++} ++ ++static void ++sd_iostats_seq_stop(struct seq_file *p, void *v) ++{ ++} ++ ++static struct seq_operations sd_iostats_seqops = { ++ .start = sd_iostats_seq_start, ++ .stop = sd_iostats_seq_stop, ++ .next = sd_iostats_seq_next, ++ .show = sd_iostats_seq_show, ++}; ++ ++static int ++sd_iostats_seq_open (struct inode *inode, struct file *file) ++{ ++ int rc; ++ ++ rc = seq_open(file, &sd_iostats_seqops); ++ if (rc != 0) ++ return rc; ++ ++ ((struct seq_file *)file->private_data)->private = PDE(inode)->data; ++ return 0; ++} ++ ++static ssize_t ++sd_iostats_seq_write(struct file *file, const char *buffer, ++ size_t len, loff_t *off) ++{ ++ struct seq_file *seq = file->private_data; ++ struct gendisk *disk = seq->private; ++ iostat_stats_t *stats = scsi_disk(disk)->stats; ++ unsigned long flags; ++ unsigned long qdepth; ++ ++ ++ spin_lock_irqsave (&stats->iostat_lock, flags); ++ qdepth = stats->iostat_queue_depth; ++ memset (stats, 0, offsetof(iostat_stats_t, iostat_lock)); ++ do_gettimeofday(&stats->iostat_timeval); ++ stats->iostat_queue_stamp = jiffies; ++ stats->iostat_queue_depth = qdepth; ++ spin_unlock_irqrestore (&stats->iostat_lock, flags); ++ ++ return len; ++} ++ ++static struct file_operations sd_iostats_proc_fops = { ++ .owner = THIS_MODULE, ++ .open = sd_iostats_seq_open, ++ .read = seq_read, ++ .write = sd_iostats_seq_write, ++ .llseek = seq_lseek, ++ .release = seq_release, ++}; ++ ++extern struct proc_dir_entry *proc_scsi; ++ ++void ++sd_iostats_init(void) ++{ ++ if (proc_scsi == NULL) { ++ printk(KERN_WARNING "No access to sd iostats: " ++ "proc_scsi is NULL\n"); ++ return; ++ } ++ ++ sd_iostats_procdir = create_proc_entry(sd_iostats_procdir_name, ++ S_IFDIR | S_IRUGO | S_IXUGO, ++ proc_scsi); ++ if (sd_iostats_procdir == NULL) { ++ printk(KERN_WARNING "No access to sd iostats: " ++ "can't create /proc/scsi/%s\n", sd_iostats_procdir_name); ++ return; ++ } ++} ++ ++void sd_iostats_fini(void) ++{ ++ if (proc_scsi != NULL && sd_iostats_procdir != NULL) ++ remove_proc_entry(sd_iostats_procdir_name, proc_scsi); ++ ++ sd_iostats_procdir = NULL; ++} ++ ++void sd_iostats_finish_req(struct scsi_cmnd *SCpnt) ++{ ++ struct request *rq = SCpnt->request; ++ iostat_stats_t *stats; ++ unsigned long *tcounter; ++ int tbucket; ++ int tmp; ++ unsigned long irqflags; ++ unsigned long i; ++ ++ stats = scsi_disk(rq->rq_disk)->stats; ++ if (stats == NULL) ++ return; ++ ++ tmp = jiffies - rq->start_time; ++ for (tbucket = 0; tmp > 1; tbucket++) ++ tmp >>= 1; ++ if (tbucket >= IOSTAT_NCOUNTERS) ++ tbucket = IOSTAT_NCOUNTERS - 1; ++ //printk("%u ticks in D to %u\n", jiffies - rq->start_time, tbucket); ++ ++ tcounter = rq_data_dir(rq) == WRITE ? ++ &stats->iostat_wtime[tbucket] : &stats->iostat_rtime[tbucket]; ++ ++ spin_lock_irqsave(&stats->iostat_lock, irqflags); ++ ++ /* update delay stats */ ++ (*tcounter)++; ++ ++ /* update queue depth stats */ ++ i = stats->iostat_queue_depth; ++ if (i >= IOSTAT_NCOUNTERS) ++ i = IOSTAT_NCOUNTERS - 1; ++ stats->iostat_queue_ticks[i] += jiffies - stats->iostat_queue_stamp; ++ stats->iostat_queue_ticks_sum += jiffies - stats->iostat_queue_stamp; ++ BUG_ON(stats->iostat_queue_depth == 0); ++ stats->iostat_queue_depth--; ++ ++ /* update seek stats. XXX: not sure about nr_sectors */ ++ stats->iostat_sectors += rq->nr_sectors; ++ stats->iostat_reqs++; ++ if (rq->sector != stats->iostat_next_sector) { ++ stats->iostat_seek_sectors += ++ rq->sector > stats->iostat_next_sector ? ++ rq->sector - stats->iostat_next_sector : ++ stats->iostat_next_sector - rq->sector; ++ stats->iostat_seeks++; ++ } ++ stats->iostat_next_sector = rq->sector + rq->nr_sectors; ++ ++ stats->iostat_queue_stamp = jiffies; ++ ++ spin_unlock_irqrestore(&stats->iostat_lock, irqflags); ++} ++ ++void sd_iostats_start_req(struct scsi_cmnd *SCpnt) ++{ ++ struct request *rq = SCpnt->request; ++ iostat_stats_t *stats; ++ iostat_counter_t *counter; ++ int bucket; ++ int tbucket; ++ int tmp; ++ unsigned long irqflags; ++ unsigned long i; ++ int nsect; ++ ++ stats = scsi_disk(rq->rq_disk)->stats; ++ if (stats == NULL) ++ return; ++ ++ nsect = scsi_bufflen(SCpnt) >> 9; ++ for (bucket = 0, tmp = nsect; tmp > 1; bucket++) ++ tmp >>= 1; ++ ++ if (bucket >= IOSTAT_NCOUNTERS) { ++ printk (KERN_ERR "sd_iostats_bump: nsect %d too big\n", nsect); ++ BUG(); ++ } ++ ++ counter = rq_data_dir(rq) == WRITE ? ++ &stats->iostat_write_histogram[bucket] : ++ &stats->iostat_read_histogram[bucket]; ++ ++ tmp = jiffies - rq->start_time; ++ for (tbucket = 0; tmp > 1; tbucket++) ++ tmp >>= 1; ++ if (tbucket >= IOSTAT_NCOUNTERS) ++ tbucket = IOSTAT_NCOUNTERS - 1; ++ //printk("%u ticks in Q to %u\n", jiffies - rq->start_time, tbucket); ++ ++ /* an ugly hack to know exact processing time. the right ++ * solution is to add one more field to struct request ++ * hopefully it will break nothing ... */ ++ rq->start_time = jiffies; ++ ++ spin_lock_irqsave(&stats->iostat_lock, irqflags); ++ ++ /* update queue depth stats */ ++ i = stats->iostat_queue_depth; ++ if (i >= IOSTAT_NCOUNTERS) ++ i = IOSTAT_NCOUNTERS - 1; ++ stats->iostat_queue_ticks[i] += jiffies - stats->iostat_queue_stamp; ++ stats->iostat_queue_ticks_sum += jiffies - stats->iostat_queue_stamp; ++ stats->iostat_queue_depth++; ++ ++ /* update delay stats */ ++ if (rq_data_dir(rq) == WRITE) { ++ stats->iostat_wtime_in_queue[tbucket]++; ++ stats->iostat_write_reqs++; ++ } else { ++ stats->iostat_rtime_in_queue[tbucket]++; ++ stats->iostat_read_reqs++; ++ } ++ ++ /* update size stats */ ++ counter->iostat_size += nsect; ++ counter->iostat_count++; ++ ++ stats->iostat_queue_stamp = jiffies; ++ ++ spin_unlock_irqrestore(&stats->iostat_lock, irqflags); ++} ++#endif ++ + /** + * scsi_disk_release - Called to free the scsi_disk structure + * @dev: pointer to embedded class device +@@ -1773,10 +2184,16 @@ + idr_remove(&sd_index_idr, sdkp->index); + spin_unlock(&sd_index_lock); + ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++ if (sdkp->stats) { ++ remove_proc_entry(disk->disk_name, sd_iostats_procdir); ++ kfree(sdkp->stats); ++ sdkp->stats = NULL; ++ } ++#endif + disk->private_data = NULL; + put_disk(disk); + put_device(&sdkp->device->sdev_gendev); +- + kfree(sdkp); + } + +@@ -1890,6 +2307,8 @@ + if (!majors) + return -ENODEV; + ++ sd_iostats_init(); ++ + err = class_register(&sd_disk_class); + if (err) + goto err_out; +@@ -1905,6 +2324,7 @@ + err_out: + for (i = 0; i < SD_MAJORS; i++) + unregister_blkdev(sd_major(i), "sd"); ++ sd_iostats_fini(); + return err; + } + +Index: linux-source-2.6.26/include/scsi/sd.h +=================================================================== +--- linux-source-2.6.26.orig/include/scsi/sd.h 2009-10-09 13:33:25.000000000 +0200 ++++ linux-source-2.6.26/include/scsi/sd.h 2009-10-09 13:39:06.000000000 +0200 +@@ -31,6 +31,46 @@ + */ + #define SD_BUF_SIZE 512 + ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++typedef struct { ++ unsigned long long iostat_size; ++ unsigned long long iostat_count; ++} iostat_counter_t; ++ ++#define IOSTAT_NCOUNTERS 16 ++typedef struct { ++ iostat_counter_t iostat_read_histogram[IOSTAT_NCOUNTERS]; ++ iostat_counter_t iostat_write_histogram[IOSTAT_NCOUNTERS]; ++ struct timeval iostat_timeval; ++ ++ /* queue depth: how well the pipe is filled up */ ++ unsigned long long iostat_queue_ticks[IOSTAT_NCOUNTERS]; ++ unsigned long long iostat_queue_ticks_sum; ++ unsigned long iostat_queue_depth; ++ unsigned long iostat_queue_stamp; ++ ++ /* seeks: how linear the traffic is */ ++ unsigned long long iostat_next_sector; ++ unsigned long long iostat_seek_sectors; ++ unsigned long long iostat_seeks; ++ unsigned long long iostat_sectors; ++ unsigned long long iostat_reqs; ++ unsigned long iostat_read_reqs; ++ unsigned long iostat_write_reqs; ++ ++ /* process time: how long it takes to process requests */ ++ unsigned long iostat_rtime[IOSTAT_NCOUNTERS]; ++ unsigned long iostat_wtime[IOSTAT_NCOUNTERS]; ++ ++ /* queue time: how long process spent in elevator's queue */ ++ unsigned long iostat_rtime_in_queue[IOSTAT_NCOUNTERS]; ++ unsigned long iostat_wtime_in_queue[IOSTAT_NCOUNTERS]; ++ ++ /* must be the last field, as it's used to know size to be memset'ed */ ++ spinlock_t iostat_lock; ++} ____cacheline_aligned_in_smp iostat_stats_t; ++#endif ++ + struct scsi_disk { + struct scsi_driver *driver; /* always &sd_template */ + struct scsi_device *device; +@@ -45,6 +85,9 @@ + unsigned WCE : 1; /* state of disk WCE bit */ + unsigned RCD : 1; /* state of disk RCD bit, unused */ + unsigned DPOFUA : 1; /* state of disk DPOFUA bit */ ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++ iostat_stats_t *stats; /* scsi disk statistics */ ++#endif + }; + #define to_scsi_disk(obj) container_of(obj,struct scsi_disk,dev) + --- lustre-1.8.3.orig/lustre/kernel_patches/patches/md-mmp-unplug-dev-2.6.27-vanilla.patch +++ lustre-1.8.3/lustre/kernel_patches/patches/md-mmp-unplug-dev-2.6.27-vanilla.patch @@ -0,0 +1,22 @@ +Index: linux-2.6.27.36/drivers/md/raid5.c +=================================================================== +--- linux-2.6.27.36.orig/drivers/md/raid5.c 2009-10-05 17:19:01.000000000 +0200 ++++ linux-2.6.27.36/drivers/md/raid5.c 2009-10-08 15:06:10.000000000 +0200 +@@ -1726,6 +1726,8 @@ + bi->bi_next = *bip; + *bip = bi; + bi->bi_phys_segments ++; ++ if (bio_sync(bi) && !forwrite) ++ clear_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); /* force to read from disk. */ + spin_unlock_irq(&conf->device_lock); + spin_unlock(&sh->lock); + +@@ -3478,6 +3480,8 @@ + + bio_endio(bi, 0); + } ++ if (bio_sync(bi)) ++ raid5_unplug_device(q); + return 0; + } + --- lustre-1.8.3.orig/lustre/kernel_patches/patches/debian-2.6.26.diff +++ lustre-1.8.3/lustre/kernel_patches/patches/debian-2.6.26.diff @@ -0,0 +1,1210 @@ +diff -u -r debian-2.6.26/Documentation/filesystems/ext2.txt debian-2.6.26_lustre.1.8.2/Documentation/filesystems/ext2.txt +--- debian-2.6.26/Documentation/filesystems/ext2.txt 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/Documentation/filesystems/ext2.txt 2010-02-12 15:11:19.000000000 +0100 +@@ -58,6 +58,22 @@ + + xip Use execute in place (no caching) if possible + ++iopen Makes an invisible pseudo-directory called ++ __iopen__ available in the root directory ++ of the filesystem. Allows open-by-inode- ++ number. i.e., inode 3145 can be accessed ++ via /mntpt/__iopen__/3145 ++ ++iopen_nopriv This option makes the iopen directory be ++ world-readable. This may be safer since it ++ allows daemons to run as an unprivileged user, ++ however it significantly changes the security ++ model of a Unix filesystem, since previously ++ all files under a mode 700 directory were not ++ generally avilable even if the ++ permissions on the file itself is ++ world-readable. ++ + grpquota,noquota,quota,usrquota Quota options are silently ignored by ext2. + + +diff -u -r debian-2.6.26/block/blk-core.c debian-2.6.26_lustre.1.8.2/block/blk-core.c +--- debian-2.6.26/block/blk-core.c 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/block/blk-core.c 2010-02-12 15:14:32.000000000 +0100 +@@ -1270,6 +1270,8 @@ + + #endif /* CONFIG_FAIL_MAKE_REQUEST */ + ++int dev_check_rdonly(struct block_device *bdev); ++ + /* + * Check whether this bio extends beyond the end of the device. + */ +@@ -1371,6 +1373,12 @@ + + if (unlikely(test_bit(QUEUE_FLAG_DEAD, &q->queue_flags))) + goto end_io; ++ /* this is cfs's dev_rdonly check */ ++ if (bio->bi_rw == WRITE && ++ dev_check_rdonly(bio->bi_bdev)) { ++ bio_endio(bio, 0); ++ break; ++ } + + if (should_fail_request(bio)) + goto end_io; +@@ -2028,6 +2036,91 @@ + } + EXPORT_SYMBOL(kblockd_flush_work); + ++ /* ++ * Debug code for turning block devices "read-only" (will discard writes ++ * silently). This is for filesystem crash/recovery testing. ++ */ ++struct deventry { ++ dev_t dev; ++ struct deventry *next; ++}; ++ ++static struct deventry *devlist = NULL; ++static spinlock_t devlock = SPIN_LOCK_UNLOCKED; ++ ++int dev_check_rdonly(struct block_device *bdev) ++{ ++ struct deventry *cur; ++ if (!bdev) return 0; ++ spin_lock(&devlock); ++ cur = devlist; ++ while(cur) { ++ if (bdev->bd_dev == cur->dev) { ++ spin_unlock(&devlock); ++ return 1; ++ } ++ cur = cur->next; ++ } ++ spin_unlock(&devlock); ++ return 0; ++} ++ ++void dev_set_rdonly(struct block_device *bdev) ++{ ++ struct deventry *newdev, *cur; ++ ++ if (!bdev) ++ return; ++ newdev = kmalloc(sizeof(struct deventry), GFP_KERNEL); ++ if (!newdev) ++ return; ++ ++ spin_lock(&devlock); ++ cur = devlist; ++ while(cur) { ++ if (bdev->bd_dev == cur->dev) { ++ spin_unlock(&devlock); ++ kfree(newdev); ++ return; ++ } ++ cur = cur->next; ++ } ++ newdev->dev = bdev->bd_dev; ++ newdev->next = devlist; ++ devlist = newdev; ++ spin_unlock(&devlock); ++ printk(KERN_WARNING "Turning device %s (%#x) read-only\n", ++ bdev->bd_disk ? bdev->bd_disk->disk_name : "", bdev->bd_dev); ++} ++ ++void dev_clear_rdonly(struct block_device *bdev) ++{ ++ struct deventry *cur, *last = NULL; ++ if (!bdev) return; ++ spin_lock(&devlock); ++ cur = devlist; ++ while(cur) { ++ if (bdev->bd_dev == cur->dev) { ++ if (last) ++ last->next = cur->next; ++ else ++ devlist = cur->next; ++ spin_unlock(&devlock); ++ kfree(cur); ++ printk(KERN_WARNING "Removing read-only on %s (%#x)\n", ++ bdev->bd_disk ? bdev->bd_disk->disk_name : ++ "unknown block", bdev->bd_dev); ++ return; ++ } ++ last = cur; ++ cur = cur->next; ++ } ++ spin_unlock(&devlock); ++} ++ ++EXPORT_SYMBOL(dev_set_rdonly); ++EXPORT_SYMBOL(dev_clear_rdonly); ++ + int __init blk_dev_init(void) + { + int i; +diff -u -r debian-2.6.26/drivers/md/raid5.c debian-2.6.26_lustre.1.8.2/drivers/md/raid5.c +--- debian-2.6.26/drivers/md/raid5.c 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/drivers/md/raid5.c 2010-02-12 15:19:25.000000000 +0100 +@@ -1817,6 +1817,8 @@ + bi->bi_next = *bip; + *bip = bi; + bi->bi_phys_segments ++; ++ if (bio_sync(bi) && !forwrite) ++ clear_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); /* force to read from disk. */ + spin_unlock_irq(&conf->device_lock); + spin_unlock(&sh->lock); + +@@ -3699,6 +3701,8 @@ + test_bit(BIO_UPTODATE, &bi->bi_flags) + ? 0 : -EIO); + } ++ if (bio_sync(bi)) ++ raid5_unplug_device(q); + return 0; + } + +diff -u -r debian-2.6.26/drivers/scsi/Kconfig debian-2.6.26_lustre.1.8.2/drivers/scsi/Kconfig +--- debian-2.6.26/drivers/scsi/Kconfig 2009-12-26 09:14:53.000000000 +0100 ++++ debian-2.6.26_lustre.1.8.2/drivers/scsi/Kconfig 2010-02-12 15:20:02.000000000 +0100 +@@ -81,6 +81,14 @@ + In this case, do not compile the driver for your SCSI host adapter + (below) as a module either. + ++config SD_IOSTATS ++ bool "Enable SCSI disk I/O stats" ++ depends on BLK_DEV_SD ++ default y ++ ---help--- ++ This enables SCSI disk I/O stats collection. You must also enable ++ /proc file system support if you want this feature. ++ + config CHR_DEV_ST + tristate "SCSI tape support" + depends on SCSI +diff -u -r debian-2.6.26/drivers/scsi/scsi_proc.c debian-2.6.26_lustre.1.8.2/drivers/scsi/scsi_proc.c +--- debian-2.6.26/drivers/scsi/scsi_proc.c 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/drivers/scsi/scsi_proc.c 2010-02-12 15:22:05.000000000 +0100 +@@ -40,7 +40,8 @@ + /* 4K page size, but our output routines, use some slack for overruns */ + #define PROC_BLOCK_SIZE (3*1024) + +-static struct proc_dir_entry *proc_scsi; ++struct proc_dir_entry *proc_scsi; ++EXPORT_SYMBOL(proc_scsi); + + /* Protect sht->present and sht->proc_dir */ + static DEFINE_MUTEX(global_host_template_mutex); +diff -u -r debian-2.6.26/drivers/scsi/sd.c debian-2.6.26_lustre.1.8.2/drivers/scsi/sd.c +--- debian-2.6.26/drivers/scsi/sd.c 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/drivers/scsi/sd.c 2010-02-17 14:13:32.000000000 +0100 +@@ -107,6 +107,24 @@ + * object after last put) */ + static DEFINE_MUTEX(sd_ref_mutex); + ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++# include ++# include ++struct proc_dir_entry *sd_iostats_procdir = NULL; ++char sd_iostats_procdir_name[] = "sd_iostats"; ++static struct file_operations sd_iostats_proc_fops; ++ ++extern void sd_iostats_init(void); ++extern void sd_iostats_fini(void); ++void sd_iostats_start_req(struct scsi_cmnd *SCpnt); ++void sd_iostats_finish_req(struct scsi_cmnd *SCpnt); ++#else ++static inline void sd_iostats_init(void) {} ++static inline void sd_iostats_fini(void) {} ++static inline void sd_iostats_start_req(struct scsi_cmnd *SCpnt) {} ++static inline void sd_iostats_finish_req(struct scsi_cmnd *SCpnt) {} ++#endif ++ + static const char *sd_cache_types[] = { + "write through", "none", "write back", + "write back, no read (daft)" +@@ -531,6 +549,8 @@ + } + SCpnt->sdb.length = this_count * sdp->sector_size; + ++ sd_iostats_start_req(SCpnt); ++ + /* + * We shouldn't disconnect in the middle of a sector, so with a dumb + * host adapter, it's safe to assume that we can at least transfer +@@ -667,7 +687,7 @@ + int diskinfo[4]; + + /* default to most commonly used values */ +- diskinfo[0] = 0x40; /* 1 << 6 */ ++ diskinfo[0] = 0x40; /* 1 << 6 */ + diskinfo[1] = 0x20; /* 1 << 5 */ + diskinfo[2] = sdkp->capacity >> 11; + +@@ -1023,6 +1043,7 @@ + break; + } + out: ++ sd_iostats_finish_req(SCpnt); + return good_bytes; + } + +@@ -1711,6 +1732,36 @@ + gd->flags = GENHD_FL_DRIVERFS; + if (sdp->removable) + gd->flags |= GENHD_FL_REMOVABLE; ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++ sdkp->stats = kzalloc(sizeof(iostat_stats_t), GFP_KERNEL); ++ if (!sdkp->stats) { ++ printk(KERN_WARNING "cannot allocate iostat structure for" ++ "%s\n", gd->disk_name); ++ } else { ++ do_gettimeofday(&sdkp->stats->iostat_timeval); ++ sdkp->stats->iostat_queue_stamp = jiffies; ++ spin_lock_init(&sdkp->stats->iostat_lock); ++ if (sd_iostats_procdir) { ++ struct proc_dir_entry *pde; ++ pde = create_proc_entry(gd->disk_name, S_IRUGO | S_IWUSR, ++ sd_iostats_procdir); ++ if (!pde) { ++ printk(KERN_WARNING "Can't create /proc/scsi/" ++ "%s/%s\n", ++ sd_iostats_procdir_name, ++ gd->disk_name); ++ kfree(sdkp->stats); ++ sdkp->stats = NULL; ++ } else { ++ pde->proc_fops = &sd_iostats_proc_fops; ++ pde->data = gd; ++ } ++ } else { ++ kfree(sdkp->stats); ++ sdkp->stats = NULL; ++ } ++ } ++#endif + + dev_set_drvdata(dev, sdkp); + add_disk(gd); +@@ -1755,6 +1806,366 @@ + return 0; + } + ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++static int ++ sd_iostats_seq_show(struct seq_file *seq, void *v) ++ { ++ struct timeval now; ++ struct gendisk *disk = seq->private; ++ iostat_stats_t *stats; ++ unsigned long long read_len; ++ unsigned long long read_len_tot; ++ unsigned long read_num; ++ unsigned long read_num_tot; ++ unsigned long long write_len; ++ unsigned long long write_len_tot; ++ unsigned long write_num; ++ unsigned long write_num_tot; ++ int i; ++ int maxi; ++ ++ stats = scsi_disk(disk)->stats; ++ if (stats == NULL) { ++ printk(KERN_ERR "sd_iostats_seq_show: NULL stats entry\n"); ++ BUG(); ++ } ++ ++ do_gettimeofday(&now); ++ now.tv_sec -= stats->iostat_timeval.tv_sec; ++ now.tv_usec -= stats->iostat_timeval.tv_usec; ++ if (now.tv_usec < 0) { ++ now.tv_usec += 1000000; ++ now.tv_sec--; ++ } ++ ++ /* this sampling races with updates */ ++ seq_printf(seq, "index: %lu snapshot_time: %lu.%06lu\n", ++ (unsigned long) scsi_disk(disk)->index, ++ now.tv_sec, now.tv_usec); ++ ++ for (i = IOSTAT_NCOUNTERS - 1; i > 0; i--) ++ if (stats->iostat_read_histogram[i].iostat_count != 0 || ++ stats->iostat_write_histogram[i].iostat_count != 0) ++ break; ++ maxi = i; ++ ++ seq_printf(seq, "%8s %8s %12s %8s %12s\n", "size", ++ "reads", "total", "writes", "total"); ++ ++ read_len_tot = write_len_tot = 0; ++ read_num_tot = write_num_tot = 0; ++ for (i = 0; i <= maxi; i++) { ++ read_len = stats->iostat_read_histogram[i].iostat_size; ++ read_len_tot += read_len; ++ read_num = stats->iostat_read_histogram[i].iostat_count; ++ read_num_tot += read_num; ++ ++ write_len = stats->iostat_write_histogram[i].iostat_size; ++ write_len_tot += write_len; ++ write_num = stats->iostat_write_histogram[i].iostat_count; ++ write_num_tot += write_num; ++ ++ seq_printf (seq, "%8d %8lu %12llu %8lu %12llu\n", ++ 512<iostat_queue_ticks[i]; ++ if (ticks == 0) ++ continue; ++ percent = stats->iostat_queue_ticks[i] * 100; ++ do_div(percent, stats->iostat_queue_ticks_sum); ++ seq_printf(seq, "%8d %8llu %8llu\n", i, ticks, percent); ++ } ++ ++ if (stats->iostat_reqs != 0) { ++ unsigned long long aveseek = 0, percent = 0; ++ ++ if (stats->iostat_seeks) { ++ aveseek = stats->iostat_seek_sectors; ++ do_div(aveseek, stats->iostat_seeks); ++ percent = stats->iostat_seeks * 100; ++ do_div(percent, stats->iostat_reqs); ++ } ++ ++ seq_printf(seq, "\n%llu sectors in %llu reqs: %llu seek(s) over " ++ "%llu sectors in ave, %llu%% of all reqs\n", ++ stats->iostat_sectors, stats->iostat_reqs, ++ stats->iostat_seeks, aveseek, percent); ++ } ++ ++ seq_printf(seq, "\n%16s %8s %8s %8s %8s\n", "process time", "reads", ++ "%%", "writes", "%%"); ++ for (i = 0; i < IOSTAT_NCOUNTERS; i++) { ++ unsigned long read_percent = 0, write_percent = 0; ++ if (stats->iostat_wtime[i] == 0 && ++ stats->iostat_rtime[i] == 0) ++ continue; ++ if (stats->iostat_read_reqs) ++ read_percent = stats->iostat_rtime[i] * 100 / ++ stats->iostat_read_reqs; ++ if (stats->iostat_write_reqs) ++ write_percent = stats->iostat_wtime[i] * 100 / ++ stats->iostat_write_reqs; ++ seq_printf(seq, "%16u %8lu %8lu %8lu %8lu\n", ++ jiffies_to_msecs(((1UL << i) >> 1) << 1), ++ stats->iostat_rtime[i], read_percent, ++ stats->iostat_wtime[i], write_percent); ++ } ++ ++ seq_printf(seq, "\n%16s %8s %8s %8s %8s\n", "time in queue", "reads", ++ "%%", "writes", "%%"); ++ for (i = 0; i < IOSTAT_NCOUNTERS; i++) { ++ unsigned long read_percent = 0, write_percent = 0; ++ if (stats->iostat_wtime_in_queue[i] == 0 && ++ stats->iostat_rtime_in_queue[i] == 0) ++ continue; ++ if (stats->iostat_read_reqs) ++ read_percent = stats->iostat_rtime_in_queue[i] * 100 / ++ stats->iostat_read_reqs; ++ if (stats->iostat_write_reqs) ++ write_percent = stats->iostat_wtime_in_queue[i] * 100 / ++ stats->iostat_write_reqs; ++ seq_printf(seq, "%16u %8lu %8lu %8lu %8lu\n", ++ jiffies_to_msecs(((1UL << i) >> 1) << 1), ++ stats->iostat_rtime_in_queue[i], ++ read_percent, ++ stats->iostat_wtime_in_queue[i], ++ write_percent); ++ } ++ ++ return 0; ++ } ++ ++ static void * ++ sd_iostats_seq_start(struct seq_file *p, loff_t *pos) ++ { ++ return (*pos == 0) ? (void *)1 : NULL; ++ } ++ ++ static void * ++ sd_iostats_seq_next(struct seq_file *p, void *v, loff_t *pos) ++ { ++ ++*pos; ++ return NULL; ++ } ++ ++ static void ++ sd_iostats_seq_stop(struct seq_file *p, void *v) ++ { ++ } ++ ++ static struct seq_operations sd_iostats_seqops = { ++ .start = sd_iostats_seq_start, ++ .stop = sd_iostats_seq_stop, ++ .next = sd_iostats_seq_next, ++ .show = sd_iostats_seq_show, ++ }; ++ ++ static int ++ sd_iostats_seq_open (struct inode *inode, struct file *file) ++ { ++ int rc; ++ ++ rc = seq_open(file, &sd_iostats_seqops); ++ if (rc != 0) ++ return rc; ++ ++ ((struct seq_file *)file->private_data)->private = PDE(inode)->data; ++ return 0; ++ } ++ ++ static ssize_t ++ sd_iostats_seq_write(struct file *file, const char *buffer, ++ size_t len, loff_t *off) ++ { ++ struct seq_file *seq = file->private_data; ++ struct gendisk *disk = seq->private; ++ iostat_stats_t *stats = scsi_disk(disk)->stats; ++ unsigned long flags; ++ unsigned long qdepth; ++ ++ ++ spin_lock_irqsave (&stats->iostat_lock, flags); ++ qdepth = stats->iostat_queue_depth; ++ memset (stats, 0, offsetof(iostat_stats_t, iostat_lock)); ++ do_gettimeofday(&stats->iostat_timeval); ++ stats->iostat_queue_stamp = jiffies; ++ stats->iostat_queue_depth = qdepth; ++ spin_unlock_irqrestore (&stats->iostat_lock, flags); ++ ++ return len; ++ } ++ ++ static struct file_operations sd_iostats_proc_fops = { ++ .owner = THIS_MODULE, ++ .open = sd_iostats_seq_open, ++ .read = seq_read, ++ .write = sd_iostats_seq_write, ++ .llseek = seq_lseek, ++ .release = seq_release, ++ }; ++ ++ extern struct proc_dir_entry *proc_scsi; ++ ++ void ++ sd_iostats_init(void) ++ { ++ if (proc_scsi == NULL) { ++ printk(KERN_WARNING "No access to sd iostats: " ++ "proc_scsi is NULL\n"); ++ return; ++ } ++ ++ sd_iostats_procdir = create_proc_entry(sd_iostats_procdir_name, ++ S_IFDIR | S_IRUGO | S_IXUGO, ++ proc_scsi); ++ if (sd_iostats_procdir == NULL) { ++ printk(KERN_WARNING "No access to sd iostats: " ++ "can't create /proc/scsi/%s\n", sd_iostats_procdir_name); ++ return; ++ } ++ } ++ ++ void sd_iostats_fini(void) ++ { ++ if (proc_scsi != NULL && sd_iostats_procdir != NULL) ++ remove_proc_entry(sd_iostats_procdir_name, proc_scsi); ++ ++ sd_iostats_procdir = NULL; ++ } ++ ++ void sd_iostats_finish_req(struct scsi_cmnd *SCpnt) ++ { ++ struct request *rq = SCpnt->request; ++ iostat_stats_t *stats; ++ unsigned long *tcounter; ++ int tbucket; ++ int tmp; ++ unsigned long irqflags; ++ unsigned long i; ++ ++ stats = scsi_disk(rq->rq_disk)->stats; ++ if (stats == NULL) ++ return; ++ ++ tmp = jiffies - rq->start_time; ++ for (tbucket = 0; tmp > 1; tbucket++) ++ tmp >>= 1; ++ if (tbucket >= IOSTAT_NCOUNTERS) ++ tbucket = IOSTAT_NCOUNTERS - 1; ++ //printk("%u ticks in D to %u\n", jiffies - rq->start_time, tbucket); ++ ++ tcounter = rq_data_dir(rq) == WRITE ? ++ &stats->iostat_wtime[tbucket] : &stats->iostat_rtime[tbucket]; ++ ++ spin_lock_irqsave(&stats->iostat_lock, irqflags); ++ ++ /* update delay stats */ ++ (*tcounter)++; ++ ++ /* update queue depth stats */ ++ i = stats->iostat_queue_depth; ++ if (i >= IOSTAT_NCOUNTERS) ++ i = IOSTAT_NCOUNTERS - 1; ++ stats->iostat_queue_ticks[i] += jiffies - stats->iostat_queue_stamp; ++ stats->iostat_queue_ticks_sum += jiffies - stats->iostat_queue_stamp; ++ BUG_ON(stats->iostat_queue_depth == 0); ++ stats->iostat_queue_depth--; ++ ++ /* update seek stats. XXX: not sure about nr_sectors */ ++ stats->iostat_sectors += rq->nr_sectors; ++ stats->iostat_reqs++; ++ if (rq->sector != stats->iostat_next_sector) { ++ stats->iostat_seek_sectors += ++ rq->sector > stats->iostat_next_sector ? ++ rq->sector - stats->iostat_next_sector : ++ stats->iostat_next_sector - rq->sector; ++ stats->iostat_seeks++; ++ } ++ stats->iostat_next_sector = rq->sector + rq->nr_sectors; ++ ++ stats->iostat_queue_stamp = jiffies; ++ ++ spin_unlock_irqrestore(&stats->iostat_lock, irqflags); ++ } ++ ++ void sd_iostats_start_req(struct scsi_cmnd *SCpnt) ++ { ++ struct request *rq = SCpnt->request; ++ iostat_stats_t *stats; ++ iostat_counter_t *counter; ++ int bucket; ++ int tbucket; ++ int tmp; ++ unsigned long irqflags; ++ unsigned long i; ++ int nsect; ++ ++ stats = scsi_disk(rq->rq_disk)->stats; ++ if (stats == NULL) ++ return; ++ ++ nsect = scsi_bufflen(SCpnt) >> 9; ++ for (bucket = 0, tmp = nsect; tmp > 1; bucket++) ++ tmp >>= 1; ++ ++ if (bucket >= IOSTAT_NCOUNTERS) { ++ printk (KERN_ERR "sd_iostats_bump: nsect %d too big\n", nsect); ++ BUG(); ++ } ++ ++ counter = rq_data_dir(rq) == WRITE ? ++ &stats->iostat_write_histogram[bucket] : ++ &stats->iostat_read_histogram[bucket]; ++ ++ tmp = jiffies - rq->start_time; ++ for (tbucket = 0; tmp > 1; tbucket++) ++ tmp >>= 1; ++ if (tbucket >= IOSTAT_NCOUNTERS) ++ tbucket = IOSTAT_NCOUNTERS - 1; ++ //printk("%u ticks in Q to %u\n", jiffies - rq->start_time, tbucket); ++ ++ /* an ugly hack to know exact processing time. the right ++ * solution is to add one more field to struct request ++ * hopefully it will break nothing ... */ ++ rq->start_time = jiffies; ++ ++ spin_lock_irqsave(&stats->iostat_lock, irqflags); ++ ++ /* update queue depth stats */ ++ i = stats->iostat_queue_depth; ++ if (i >= IOSTAT_NCOUNTERS) ++ i = IOSTAT_NCOUNTERS - 1; ++ stats->iostat_queue_ticks[i] += jiffies - stats->iostat_queue_stamp; ++ stats->iostat_queue_ticks_sum += jiffies - stats->iostat_queue_stamp; ++ stats->iostat_queue_depth++; ++ ++ /* update delay stats */ ++ if (rq_data_dir(rq) == WRITE) { ++ stats->iostat_wtime_in_queue[tbucket]++; ++ stats->iostat_write_reqs++; ++ } else { ++ stats->iostat_rtime_in_queue[tbucket]++; ++ stats->iostat_read_reqs++; ++ } ++ ++ /* update size stats */ ++ counter->iostat_size += nsect; ++ counter->iostat_count++; ++ ++ stats->iostat_queue_stamp = jiffies; ++ ++ spin_unlock_irqrestore(&stats->iostat_lock, irqflags); ++} ++#endif ++ + /** + * scsi_disk_release - Called to free the scsi_disk structure + * @dev: pointer to embedded class device +@@ -1773,6 +2184,13 @@ + idr_remove(&sd_index_idr, sdkp->index); + spin_unlock(&sd_index_lock); + ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++ if (sdkp->stats) { ++ remove_proc_entry(disk->disk_name, sd_iostats_procdir); ++ kfree(sdkp->stats); ++ sdkp->stats = NULL; ++ } ++#endif + disk->private_data = NULL; + put_disk(disk); + put_device(&sdkp->device->sdev_gendev); +@@ -1890,6 +2308,8 @@ + if (!majors) + return -ENODEV; + ++ sd_iostats_init(); ++ + err = class_register(&sd_disk_class); + if (err) + goto err_out; +@@ -1905,6 +2325,7 @@ + err_out: + for (i = 0; i < SD_MAJORS; i++) + unregister_blkdev(sd_major(i), "sd"); ++ sd_iostats_fini(); + return err; + } + +diff -u -r debian-2.6.26/fs/block_dev.c debian-2.6.26_lustre.1.8.2/fs/block_dev.c +--- debian-2.6.26/fs/block_dev.c 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/fs/block_dev.c 2010-02-17 14:19:50.000000000 +0100 +@@ -1125,6 +1125,7 @@ + if (bdev != bdev->bd_contains) + victim = bdev->bd_contains; + bdev->bd_contains = NULL; ++ dev_clear_rdonly(bdev); + } + unlock_kernel(); + mutex_unlock(&bdev->bd_mutex); +diff -u -r debian-2.6.26/fs/dcache.c debian-2.6.26_lustre.1.8.2/fs/dcache.c +--- debian-2.6.26/fs/dcache.c 2009-12-26 09:14:56.000000000 +0100 ++++ debian-2.6.26_lustre.1.8.2/fs/dcache.c 2010-02-17 14:24:24.000000000 +0100 +@@ -250,6 +250,13 @@ + spin_unlock(&dcache_lock); + return 0; + } ++ ++ /* network invalidation by Lustre */ ++ if (dentry->d_flags & DCACHE_LUSTRE_INVALID) { ++ spin_unlock(&dcache_lock); ++ return 0; ++ } ++ + /* + * Check whether to do a partial shrink_dcache + * to get rid of unused child entries. +@@ -1427,14 +1434,23 @@ + * + * Adds a dentry to the hash according to its name. + */ +- ++ ++void d_rehash_cond(struct dentry * entry, int lock) ++{ ++ if (lock) ++ spin_lock(&dcache_lock); ++ spin_lock(&entry->d_lock); ++ _d_rehash(entry); ++ spin_unlock(&entry->d_lock); ++ if (lock) ++ spin_unlock(&dcache_lock); ++} ++ ++EXPORT_SYMBOL(d_rehash_cond); ++ + void d_rehash(struct dentry * entry) + { +- spin_lock(&dcache_lock); +- spin_lock(&entry->d_lock); +- _d_rehash(entry); +- spin_unlock(&entry->d_lock); +- spin_unlock(&dcache_lock); ++ d_rehash_cond(entry, 1); + } + + #define do_switch(x,y) do { \ +@@ -1510,7 +1526,7 @@ + * Update the dcache to reflect the move of a file name. Negative + * dcache entries should not be moved in this way. + */ +-static void d_move_locked(struct dentry * dentry, struct dentry * target) ++void d_move_locked(struct dentry * dentry, struct dentry * target) + { + struct hlist_head *list; + +@@ -1568,6 +1584,7 @@ + spin_unlock(&dentry->d_lock); + write_sequnlock(&rename_lock); + } ++EXPORT_SYMBOL(d_move_locked); + + /** + * d_move - move a dentry +@@ -2051,6 +2068,7 @@ + + return result; + } ++EXPORT_SYMBOL(is_subdir); + + void d_genocide(struct dentry *root) + { +diff -u -r debian-2.6.26/fs/filesystems.c debian-2.6.26_lustre.1.8.2/fs/filesystems.c +--- debian-2.6.26/fs/filesystems.c 2009-12-26 09:14:56.000000000 +0100 ++++ debian-2.6.26_lustre.1.8.2/fs/filesystems.c 2010-02-17 14:26:59.000000000 +0100 +@@ -28,7 +28,9 @@ + */ + + static struct file_system_type *file_systems; +-static DEFINE_RWLOCK(file_systems_lock); ++DEFINE_RWLOCK(file_systems_lock); ++ ++EXPORT_SYMBOL(file_systems_lock); + + /* WARNING: This can be used only if we _already_ own a reference */ + void get_filesystem(struct file_system_type *fs) +diff -u -r debian-2.6.26/fs/inode.c debian-2.6.26_lustre.1.8.2/fs/inode.c +--- debian-2.6.26/fs/inode.c 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/fs/inode.c 2010-02-17 14:27:51.000000000 +0100 +@@ -409,7 +409,9 @@ + int nr_scanned; + unsigned long reap = 0; + +- mutex_lock(&iprune_mutex); ++ if (!mutex_trylock(&iprune_mutex)) ++ return; ++ + spin_lock(&inode_lock); + for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) { + struct inode *inode; +diff -u -r debian-2.6.26/fs/jbd2/checkpoint.c debian-2.6.26_lustre.1.8.2/fs/jbd2/checkpoint.c +--- debian-2.6.26/fs/jbd2/checkpoint.c 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/fs/jbd2/checkpoint.c 2010-02-17 14:28:49.000000000 +0100 +@@ -696,6 +696,7 @@ + J_ASSERT(transaction->t_checkpoint_list == NULL); + J_ASSERT(transaction->t_checkpoint_io_list == NULL); + J_ASSERT(transaction->t_updates == 0); ++ J_ASSERT(list_empty(&transaction->t_jcb)); + J_ASSERT(journal->j_committing_transaction != transaction); + J_ASSERT(journal->j_running_transaction != transaction); + +diff -u -r debian-2.6.26/fs/jbd2/commit.c debian-2.6.26_lustre.1.8.2/fs/jbd2/commit.c +--- debian-2.6.26/fs/jbd2/commit.c 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/fs/jbd2/commit.c 2010-02-17 14:29:44.000000000 +0100 +@@ -874,6 +874,30 @@ + transaction can be removed from any checkpoint list it was on + before. */ + ++ /* ++ * Call any callbacks that had been registered for handles in this ++ * transaction. It is up to the callback to free any allocated ++ * memory. ++ * ++ * The spinlocking (t_jcb_lock) here is surely unnecessary... ++ */ ++ spin_lock(&commit_transaction->t_jcb_lock); ++ if (!list_empty(&commit_transaction->t_jcb)) { ++ struct list_head *p, *n; ++ int error = is_journal_aborted(journal); ++ ++ list_for_each_safe(p, n, &commit_transaction->t_jcb) { ++ struct journal_callback *jcb; ++ ++ jcb = list_entry(p, struct journal_callback, jcb_list); ++ list_del(p); ++ spin_unlock(&commit_transaction->t_jcb_lock); ++ jcb->jcb_func(jcb, error); ++ spin_lock(&commit_transaction->t_jcb_lock); ++ } ++ } ++ spin_unlock(&commit_transaction->t_jcb_lock); ++ + jbd_debug(3, "JBD: commit phase 7\n"); + + J_ASSERT(commit_transaction->t_sync_datalist == NULL); +diff -u -r debian-2.6.26/fs/jbd2/journal.c debian-2.6.26_lustre.1.8.2/fs/jbd2/journal.c +--- debian-2.6.26/fs/jbd2/journal.c 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/fs/jbd2/journal.c 2010-02-17 14:38:15.000000000 +0100 +@@ -82,6 +82,8 @@ + EXPORT_SYMBOL(jbd2_journal_invalidatepage); + EXPORT_SYMBOL(jbd2_journal_try_to_free_buffers); + EXPORT_SYMBOL(jbd2_journal_force_commit); ++EXPORT_SYMBOL(jbd2_journal_callback_set); ++EXPORT_SYMBOL(jbd2_journal_bmap); + + static int journal_convert_superblock_v1(journal_t *, journal_superblock_t *); + static void __journal_abort_soft (journal_t *journal, int errno); +@@ -460,6 +462,7 @@ + spin_unlock(&journal->j_state_lock); + return ret; + } ++EXPORT_SYMBOL(jbd2_log_start_commit); + + /* + * Force and wait upon a commit if the calling process is not within +diff -u -r debian-2.6.26/fs/jbd2/transaction.c debian-2.6.26_lustre.1.8.2/fs/jbd2/transaction.c +--- debian-2.6.26/fs/jbd2/transaction.c 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/fs/jbd2/transaction.c 2010-02-17 14:42:20.000000000 +0100 +@@ -51,10 +51,12 @@ + transaction->t_state = T_RUNNING; + transaction->t_tid = journal->j_transaction_sequence++; + transaction->t_expires = jiffies + journal->j_commit_interval; ++ INIT_LIST_HEAD(&transaction->t_jcb); + spin_lock_init(&transaction->t_handle_lock); ++ spin_lock_init(&transaction->t_jcb_lock); + + /* Set up the commit timer for the new transaction. */ +- journal->j_commit_timer.expires = round_jiffies(transaction->t_expires); ++ journal->j_commit_timer.expires = transaction->t_expires; + add_timer(&journal->j_commit_timer); + + J_ASSERT(journal->j_running_transaction == NULL); +@@ -252,6 +254,7 @@ + memset(handle, 0, sizeof(*handle)); + handle->h_buffer_credits = nblocks; + handle->h_ref = 1; ++ INIT_LIST_HEAD(&handle->h_jcb); + + lockdep_init_map(&handle->h_lockdep_map, "jbd2_handle", + &jbd2_handle_key, 0); +@@ -1350,6 +1353,36 @@ + } + + /** ++ * void jbd2_journal_callback_set() - Register a callback function for this handle. ++ * @handle: handle to attach the callback to. ++ * @func: function to callback. ++ * @jcb: structure with additional information required by func() , and ++ * some space for jbd2 internal information. ++ * ++ * The function will be ++ * called when the transaction that this handle is part of has been ++ * committed to disk with the original callback data struct and the ++ * error status of the journal as parameters. There is no guarantee of ++ * ordering between handles within a single transaction, nor between ++ * callbacks registered on the same handle. ++ * ++ * The caller is responsible for allocating the journal_callback struct. ++ * This is to allow the caller to add as much extra data to the callback ++ * as needed, but reduce the overhead of multiple allocations. The caller ++ * allocated struct must start with a struct journal_callback at offset 0, ++ * and has the caller-specific data afterwards. ++ */ ++void jbd2_journal_callback_set(handle_t *handle, ++ void (*func)(struct journal_callback *jcb, int error), ++ struct journal_callback *jcb) ++{ ++ spin_lock(&handle->h_transaction->t_jcb_lock); ++ list_add_tail(&jcb->jcb_list, &handle->h_jcb); ++ spin_unlock(&handle->h_transaction->t_jcb_lock); ++ jcb->jcb_func = func; ++} ++ ++/** + * int jbd2_journal_stop() - complete a transaction + * @handle: tranaction to complete. + * +@@ -1423,6 +1456,11 @@ + wake_up(&journal->j_wait_transaction_locked); + } + ++ /* Move callbacks from the handle to the transaction. */ ++ spin_lock(&transaction->t_jcb_lock); ++ list_splice(&handle->h_jcb, &transaction->t_jcb); ++ spin_unlock(&transaction->t_jcb_lock); ++ + /* + * If the handle is marked SYNC, we need to set another commit + * going! We also want to force a commit if the current +diff -u -r debian-2.6.26/fs/namespace.c debian-2.6.26_lustre.1.8.2/fs/namespace.c +--- debian-2.6.26/fs/namespace.c 2009-12-26 09:14:56.000000000 +0100 ++++ debian-2.6.26_lustre.1.8.2/fs/namespace.c 2010-02-17 14:43:07.000000000 +0100 +@@ -1660,6 +1660,7 @@ + + return do_add_mount(mnt, nd, mnt_flags, NULL); + } ++EXPORT_SYMBOL(set_fs_pwd); + + /* + * add a mount into a namespace's mount tree +diff -u -r debian-2.6.26/include/linux/blkdev.h debian-2.6.26_lustre.1.8.2/include/linux/blkdev.h +--- debian-2.6.26/include/linux/blkdev.h 2009-12-26 09:14:55.000000000 +0100 ++++ debian-2.6.26_lustre.1.8.2/include/linux/blkdev.h 2010-02-18 07:44:31.000000000 +0100 +@@ -806,7 +806,7 @@ + #define MAX_PHYS_SEGMENTS 128 + #define MAX_HW_SEGMENTS 128 + #define SAFE_MAX_SECTORS 255 +-#define BLK_DEF_MAX_SECTORS 1024 ++#define BLK_DEF_MAX_SECTORS 2048 + + #define MAX_SEGMENT_SIZE 65536 + +diff -u -r debian-2.6.26/include/linux/dcache.h debian-2.6.26_lustre.1.8.2/include/linux/dcache.h +--- debian-2.6.26/include/linux/dcache.h 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/include/linux/dcache.h 2010-02-18 07:46:08.000000000 +0100 +@@ -173,6 +173,7 @@ + + #define DCACHE_REFERENCED 0x0008 /* Recently used, don't discard. */ + #define DCACHE_UNHASHED 0x0010 ++#define DCACHE_LUSTRE_INVALID 0x0040 /* Lustre invalidated */ + + #define DCACHE_INOTIFY_PARENT_WATCHED 0x0020 /* Parent inode is watched */ + +@@ -250,6 +251,7 @@ + * This adds the entry to the hash queues. + */ + extern void d_rehash(struct dentry *); ++extern void d_rehash_cond(struct dentry *, int lock); + + /** + * d_add - add dentry to hash queues +@@ -285,6 +287,7 @@ + + /* used for rename() and baskets */ + extern void d_move(struct dentry *, struct dentry *); ++extern void d_move_locked(struct dentry *, struct dentry *); + + /* appendix may either be NULL or be used for transname suffixes */ + extern struct dentry * d_lookup(struct dentry *, struct qstr *); +diff -u -r debian-2.6.26/include/linux/fs.h debian-2.6.26_lustre.1.8.2/include/linux/fs.h +--- debian-2.6.26/include/linux/fs.h 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/include/linux/fs.h 2010-02-18 07:57:56.000000000 +0100 +@@ -1832,6 +1832,10 @@ + extern void submit_bio(int, struct bio *); + extern int bdev_read_only(struct block_device *); + #endif ++#define HAVE_CLEAR_RDONLY_ON_PUT ++extern void dev_set_rdonly(struct block_device *bdev); ++extern int dev_check_rdonly(struct block_device *bdev); ++extern void dev_clear_rdonly(struct block_device *bdev); + extern int set_blocksize(struct block_device *, int); + extern int sb_set_blocksize(struct super_block *, int); + extern int sb_min_blocksize(struct super_block *, int); +@@ -1930,6 +1934,7 @@ + + extern const struct file_operations generic_ro_fops; + ++extern rwlock_t file_systems_lock; + #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) + + extern int vfs_readlink(struct dentry *, char __user *, int, const char *); +diff -u -r debian-2.6.26/include/linux/jbd2.h debian-2.6.26_lustre.1.8.2/include/linux/jbd2.h +--- debian-2.6.26/include/linux/jbd2.h 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/include/linux/jbd2.h 2010-02-18 08:08:30.000000000 +0100 +@@ -379,6 +379,27 @@ + bit_spin_unlock(BH_JournalHead, &bh->b_state); + } + ++#define HAVE_JOURNAL_CALLBACK_STATUS ++/** ++ * struct journal_callback - Base structure for callback information. ++ * @jcb_list: list information for other callbacks attached to the same handle. ++ * @jcb_func: Function to call with this callback structure. ++ * ++ * This struct is a 'seed' structure for a using with your own callback ++ * structs. If you are using callbacks you must allocate one of these ++ * or another struct of your own definition which has this struct ++ * as it's first element and pass it to journal_callback_set(). ++ * ++ * This is used internally by jbd2 to maintain callback information. ++ * ++ * See journal_callback_set for more information. ++ **/ ++struct journal_callback { ++ struct list_head jcb_list; /* t_jcb_lock */ ++ void (*jcb_func)(struct journal_callback *jcb, int error); ++ /* user data goes here */ ++}; ++ + struct jbd2_revoke_table_s; + + /** +@@ -387,6 +408,7 @@ + * @h_transaction: Which compound transaction is this update a part of? + * @h_buffer_credits: Number of remaining buffers we are allowed to dirty. + * @h_ref: Reference count on this handle ++ * @h_jcb: List of application registered callbacks for this handle. + * @h_err: Field for caller's use to track errors through large fs operations + * @h_sync: flag for sync-on-close + * @h_jdata: flag to force data journaling +@@ -412,6 +434,13 @@ + /* operations */ + int h_err; + ++ /* ++ * List of application registered callbacks for this handle. The ++ * function(s) will be called after the transaction that this handle is ++ * part of has been committed to disk. [t_jcb_lock] ++ */ ++ struct list_head h_jcb; ++ + /* Flags [no locking] */ + unsigned int h_sync: 1; /* sync-on-close */ + unsigned int h_jdata: 1; /* force data journaling */ +@@ -467,6 +496,9 @@ + * j_state_lock + * ->j_list_lock (journal_unmap_buffer) + * ++ * t_handle_lock ++ * ->t_jcb_lock ++ * + */ + + struct transaction_s +@@ -613,6 +645,15 @@ + */ + int t_handle_count; + ++ /* ++ * Protects the callback list ++ */ ++ spinlock_t t_jcb_lock; ++ /* ++ * List of registered callback functions for this transaction. ++ * Called when the transaction is committed. [t_jcb_lock] ++ */ ++ struct list_head t_jcb; + }; + + struct transaction_run_stats_s { +@@ -1016,6 +1057,9 @@ + extern int jbd2_journal_flush (journal_t *); + extern void jbd2_journal_lock_updates (journal_t *); + extern void jbd2_journal_unlock_updates (journal_t *); ++extern void jbd2_journal_callback_set(handle_t *handle, ++ void (*fn)(struct journal_callback *,int), ++ struct journal_callback *jcb); + + extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, + struct block_device *fs_dev, +diff -u -r debian-2.6.26/include/linux/mm.h debian-2.6.26_lustre.1.8.2/include/linux/mm.h +--- debian-2.6.26/include/linux/mm.h 2009-12-26 09:14:57.000000000 +0100 ++++ debian-2.6.26_lustre.1.8.2/include/linux/mm.h 2010-02-18 08:10:11.000000000 +0100 +@@ -564,6 +564,8 @@ + { + return __va(page_to_pfn(page) << PAGE_SHIFT); + } ++/* truncate.c */ ++extern void truncate_complete_page(struct address_space *mapping,struct page *); + + #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) + #define HASHED_PAGE_VIRTUAL +diff -u -r debian-2.6.26/include/scsi/sd.h debian-2.6.26_lustre.1.8.2/include/scsi/sd.h +--- debian-2.6.26/include/scsi/sd.h 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/include/scsi/sd.h 2010-02-17 14:18:52.000000000 +0100 +@@ -31,6 +31,47 @@ + */ + #define SD_BUF_SIZE 512 + ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++typedef struct { ++ unsigned long long iostat_size; ++ unsigned long long iostat_count; ++} iostat_counter_t; ++ ++#define IOSTAT_NCOUNTERS 16 ++typedef struct { ++ iostat_counter_t iostat_read_histogram[IOSTAT_NCOUNTERS]; ++ iostat_counter_t iostat_write_histogram[IOSTAT_NCOUNTERS]; ++ struct timeval iostat_timeval; ++ ++ /* queue depth: how well the pipe is filled up */ ++ unsigned long long iostat_queue_ticks[IOSTAT_NCOUNTERS]; ++ unsigned long long iostat_queue_ticks_sum; ++ unsigned long iostat_queue_depth; ++ unsigned long iostat_queue_stamp; ++ ++ /* seeks: how linear the traffic is */ ++ unsigned long long iostat_next_sector; ++ unsigned long long iostat_seek_sectors; ++ unsigned long long iostat_seeks; ++ unsigned long long iostat_sectors; ++ unsigned long long iostat_reqs; ++ unsigned long iostat_read_reqs; ++ unsigned long iostat_write_reqs; ++ ++ /* process time: how long it takes to process requests */ ++ unsigned long iostat_rtime[IOSTAT_NCOUNTERS]; ++ unsigned long iostat_wtime[IOSTAT_NCOUNTERS]; ++ ++ /* queue time: how long process spent in elevator's queue */ ++ unsigned long iostat_rtime_in_queue[IOSTAT_NCOUNTERS]; ++ unsigned long iostat_wtime_in_queue[IOSTAT_NCOUNTERS]; ++ ++ /* must be the last field, as it's used to know size to be memset'ed */ ++ spinlock_t iostat_lock; ++} ____cacheline_aligned_in_smp iostat_stats_t; ++#endif ++ ++ + struct scsi_disk { + struct scsi_driver *driver; /* always &sd_template */ + struct scsi_device *device; +@@ -45,6 +86,9 @@ + unsigned WCE : 1; /* state of disk WCE bit */ + unsigned RCD : 1; /* state of disk RCD bit, unused */ + unsigned DPOFUA : 1; /* state of disk DPOFUA bit */ ++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS)) ++ iostat_stats_t *stats; /* scsi disk statistics */ ++#endif + }; + #define to_scsi_disk(obj) container_of(obj,struct scsi_disk,dev) + +diff -u -r debian-2.6.26/kernel/sched.c debian-2.6.26_lustre.1.8.2/kernel/sched.c +--- debian-2.6.26/kernel/sched.c 2009-12-26 09:14:56.000000000 +0100 ++++ debian-2.6.26_lustre.1.8.2/kernel/sched.c 2010-02-18 08:11:02.000000000 +0100 +@@ -5477,6 +5477,7 @@ + + show_stack(p, NULL); + } ++EXPORT_SYMBOL(sched_show_task); + + void show_state_filter(unsigned long state_filter) + { +diff -u -r debian-2.6.26/mm/truncate.c debian-2.6.26_lustre.1.8.2/mm/truncate.c +--- debian-2.6.26/mm/truncate.c 2008-07-13 23:51:29.000000000 +0200 ++++ debian-2.6.26_lustre.1.8.2/mm/truncate.c 2010-02-18 08:12:58.000000000 +0100 +@@ -92,7 +92,7 @@ + * its lock, b) when a concurrent invalidate_mapping_pages got there first and + * c) when tmpfs swizzles a page between a tmpfs inode and swapper_space. + */ +-static void ++void + truncate_complete_page(struct address_space *mapping, struct page *page) + { + if (page->mapping != mapping) +@@ -108,6 +108,7 @@ + ClearPageMappedToDisk(page); + page_cache_release(page); /* pagecache ref */ + } ++EXPORT_SYMBOL_GPL(truncate_complete_page); + + /* + * This is for invalidate_mapping_pages(). That function can be called at +diff -u -r debian-2.6.26/security/security.c debian-2.6.26_lustre.1.8.2/security/security.c +--- debian-2.6.26/security/security.c 2009-12-26 09:14:57.000000000 +0100 ++++ debian-2.6.26_lustre.1.8.2/security/security.c 2010-02-18 08:13:38.000000000 +0100 +@@ -406,6 +406,7 @@ + return 0; + return security_ops->inode_unlink(dir, dentry); + } ++EXPORT_SYMBOL(security_inode_unlink); + + int security_inode_symlink(struct inode *dir, struct dentry *dentry, + const char *old_name) --- lustre-1.8.3.orig/lustre/kernel_patches/patches/quota-Increase-size-of-variables-for-limits-and-ino.patch +++ lustre-1.8.3/lustre/kernel_patches/patches/quota-Increase-size-of-variables-for-limits-and-ino.patch @@ -0,0 +1,359 @@ +From: Jan Kara +References: fate#302681 +Subject: [PATCH 02/28] quota: Increase size of variables for limits and inode usage +Patch-mainline: 2.6.29? + +So far quota was fine with quota block limits and inode limits/numbers in +a 32-bit type. Now with rapid increase in storage sizes there are coming +requests to be able to handle quota limits above 4TB / more that 2^32 inodes. +So bump up sizes of types in mem_dqblk structure to 64-bits to be able to +handle this. Also update inode allocation / checking functions to use qsize_t +and make global structure keep quota limits in bytes so that things are +consistent. + +Signed-off-by: Jan Kara +--- + fs/dquot.c | 50 ++++++++++++++++++++++++++------------------- + fs/quota_v1.c | 25 +++++++++++++++++----- + fs/quota_v2.c | 21 +++++++++++++++--- + include/linux/quota.h | 28 +++++++++++-------------- + include/linux/quotaops.h | 4 +- + 5 files changed, 79 insertions(+), 49 deletions(-) + +diff --git a/fs/dquot.c b/fs/dquot.c +index e1dac3e..758bf4a 100644 +--- a/fs/dquot.c ++++ b/fs/dquot.c +@@ -833,7 +833,7 @@ static void drop_dquot_ref(struct super_block *sb, int type) + } + } + +-static inline void dquot_incr_inodes(struct dquot *dquot, unsigned long number) ++static inline void dquot_incr_inodes(struct dquot *dquot, qsize_t number) + { + dquot->dq_dqb.dqb_curinodes += number; + } +@@ -843,7 +843,7 @@ static inline void dquot_incr_space(struct dquot *dquot, qsize_t number) + dquot->dq_dqb.dqb_curspace += number; + } + +-static inline void dquot_decr_inodes(struct dquot *dquot, unsigned long number) ++static inline void dquot_decr_inodes(struct dquot *dquot, qsize_t number) + { + if (dquot->dq_dqb.dqb_curinodes > number) + dquot->dq_dqb.dqb_curinodes -= number; +@@ -860,7 +860,7 @@ static inline void dquot_decr_space(struct dquot *dquot, qsize_t number) + dquot->dq_dqb.dqb_curspace -= number; + else + dquot->dq_dqb.dqb_curspace = 0; +- if (toqb(dquot->dq_dqb.dqb_curspace) <= dquot->dq_dqb.dqb_bsoftlimit) ++ if (dquot->dq_dqb.dqb_curspace <= dquot->dq_dqb.dqb_bsoftlimit) + dquot->dq_dqb.dqb_btime = (time_t) 0; + clear_bit(DQ_BLKS_B, &dquot->dq_flags); + } +@@ -1038,7 +1038,7 @@ static inline char ignore_hardlimit(struct dquot *dquot) + } + + /* needs dq_data_lock */ +-static int check_idq(struct dquot *dquot, ulong inodes, char *warntype) ++static int check_idq(struct dquot *dquot, qsize_t inodes, char *warntype) + { + *warntype = QUOTA_NL_NOWARN; + if (inodes <= 0 || test_bit(DQ_FAKE_B, &dquot->dq_flags)) +@@ -1077,7 +1077,7 @@ static int check_bdq(struct dquot *dquot, qsize_t space, int prealloc, char *war + return QUOTA_OK; + + if (dquot->dq_dqb.dqb_bhardlimit && +- toqb(dquot->dq_dqb.dqb_curspace + space) > dquot->dq_dqb.dqb_bhardlimit && ++ dquot->dq_dqb.dqb_curspace + space > dquot->dq_dqb.dqb_bhardlimit && + !ignore_hardlimit(dquot)) { + if (!prealloc) + *warntype = QUOTA_NL_BHARDWARN; +@@ -1085,7 +1085,7 @@ static int check_bdq(struct dquot *dquot, qsize_t space, int prealloc, char *war + } + + if (dquot->dq_dqb.dqb_bsoftlimit && +- toqb(dquot->dq_dqb.dqb_curspace + space) > dquot->dq_dqb.dqb_bsoftlimit && ++ dquot->dq_dqb.dqb_curspace + space > dquot->dq_dqb.dqb_bsoftlimit && + dquot->dq_dqb.dqb_btime && get_seconds() >= dquot->dq_dqb.dqb_btime && + !ignore_hardlimit(dquot)) { + if (!prealloc) +@@ -1094,7 +1094,7 @@ static int check_bdq(struct dquot *dquot, qsize_t space, int prealloc, char *war + } + + if (dquot->dq_dqb.dqb_bsoftlimit && +- toqb(dquot->dq_dqb.dqb_curspace + space) > dquot->dq_dqb.dqb_bsoftlimit && ++ dquot->dq_dqb.dqb_curspace + space > dquot->dq_dqb.dqb_bsoftlimit && + dquot->dq_dqb.dqb_btime == 0) { + if (!prealloc) { + *warntype = QUOTA_NL_BSOFTWARN; +@@ -1111,7 +1111,7 @@ static int check_bdq(struct dquot *dquot, qsize_t space, int prealloc, char *war + return QUOTA_OK; + } + +-static int info_idq_free(struct dquot *dquot, ulong inodes) ++static int info_idq_free(struct dquot *dquot, qsize_t inodes) + { + if (test_bit(DQ_FAKE_B, &dquot->dq_flags) || + dquot->dq_dqb.dqb_curinodes <= dquot->dq_dqb.dqb_isoftlimit) +@@ -1128,15 +1128,13 @@ static int info_idq_free(struct dquot *dquot, ulong inodes) + static int info_bdq_free(struct dquot *dquot, qsize_t space) + { + if (test_bit(DQ_FAKE_B, &dquot->dq_flags) || +- toqb(dquot->dq_dqb.dqb_curspace) <= dquot->dq_dqb.dqb_bsoftlimit) ++ dquot->dq_dqb.dqb_curspace <= dquot->dq_dqb.dqb_bsoftlimit) + return QUOTA_NL_NOWARN; + +- if (toqb(dquot->dq_dqb.dqb_curspace - space) <= +- dquot->dq_dqb.dqb_bsoftlimit) ++ if (dquot->dq_dqb.dqb_curspace - space <= dquot->dq_dqb.dqb_bsoftlimit) + return QUOTA_NL_BSOFTBELOW; +- if (toqb(dquot->dq_dqb.dqb_curspace) >= dquot->dq_dqb.dqb_bhardlimit && +- toqb(dquot->dq_dqb.dqb_curspace - space) < +- dquot->dq_dqb.dqb_bhardlimit) ++ if (dquot->dq_dqb.dqb_curspace >= dquot->dq_dqb.dqb_bhardlimit && ++ dquot->dq_dqb.dqb_curspace - space < dquot->dq_dqb.dqb_bhardlimit) + return QUOTA_NL_BHARDBELOW; + return QUOTA_NL_NOWARN; + } +@@ -1279,7 +1277,7 @@ warn_put_all: + /* + * This operation can block, but only after everything is updated + */ +-int dquot_alloc_inode(const struct inode *inode, unsigned long number) ++int dquot_alloc_inode(const struct inode *inode, qsize_t number) + { + int cnt, ret = NO_QUOTA; + char warntype[MAXQUOTAS]; +@@ -1364,7 +1362,7 @@ out_sub: + /* + * This operation can block, but only after everything is updated + */ +-int dquot_free_inode(const struct inode *inode, unsigned long number) ++int dquot_free_inode(const struct inode *inode, qsize_t number) + { + unsigned int cnt; + char warntype[MAXQUOTAS]; +@@ -1881,14 +1879,24 @@ int vfs_dq_quota_on_remount(struct super_block *sb) + return ret; + } + ++static inline qsize_t qbtos(qsize_t blocks) ++{ ++ return blocks << QIF_DQBLKSIZE_BITS; ++} ++ ++static inline qsize_t stoqb(qsize_t space) ++{ ++ return (space + QIF_DQBLKSIZE - 1) >> QIF_DQBLKSIZE_BITS; ++} ++ + /* Generic routine for getting common part of quota structure */ + static void do_get_dqblk(struct dquot *dquot, struct if_dqblk *di) + { + struct mem_dqblk *dm = &dquot->dq_dqb; + + spin_lock(&dq_data_lock); +- di->dqb_bhardlimit = dm->dqb_bhardlimit; +- di->dqb_bsoftlimit = dm->dqb_bsoftlimit; ++ di->dqb_bhardlimit = stoqb(dm->dqb_bhardlimit); ++ di->dqb_bsoftlimit = stoqb(dm->dqb_bsoftlimit); + di->dqb_curspace = dm->dqb_curspace; + di->dqb_ihardlimit = dm->dqb_ihardlimit; + di->dqb_isoftlimit = dm->dqb_isoftlimit; +@@ -1935,8 +1943,8 @@ static int do_set_dqblk(struct dquot *dquot, struct if_dqblk *di) + check_blim = 1; + } + if (di->dqb_valid & QIF_BLIMITS) { +- dm->dqb_bsoftlimit = di->dqb_bsoftlimit; +- dm->dqb_bhardlimit = di->dqb_bhardlimit; ++ dm->dqb_bsoftlimit = qbtos(di->dqb_bsoftlimit); ++ dm->dqb_bhardlimit = qbtos(di->dqb_bhardlimit); + check_blim = 1; + } + if (di->dqb_valid & QIF_INODES) { +@@ -1954,7 +1962,7 @@ static int do_set_dqblk(struct dquot *dquot, struct if_dqblk *di) + dm->dqb_itime = di->dqb_itime; + + if (check_blim) { +- if (!dm->dqb_bsoftlimit || toqb(dm->dqb_curspace) < dm->dqb_bsoftlimit) { ++ if (!dm->dqb_bsoftlimit || dm->dqb_curspace < dm->dqb_bsoftlimit) { + dm->dqb_btime = 0; + clear_bit(DQ_BLKS_B, &dquot->dq_flags); + } +diff --git a/fs/quota_v1.c b/fs/quota_v1.c +index 5ae15b1..3e078ee 100644 +--- a/fs/quota_v1.c ++++ b/fs/quota_v1.c +@@ -14,14 +14,27 @@ MODULE_AUTHOR("Jan Kara"); + MODULE_DESCRIPTION("Old quota format support"); + MODULE_LICENSE("GPL"); + ++#define QUOTABLOCK_BITS 10 ++#define QUOTABLOCK_SIZE (1 << QUOTABLOCK_BITS) ++ ++static inline qsize_t v1_stoqb(qsize_t space) ++{ ++ return (space + QUOTABLOCK_SIZE - 1) >> QUOTABLOCK_BITS; ++} ++ ++static inline qsize_t v1_qbtos(qsize_t blocks) ++{ ++ return blocks << QUOTABLOCK_BITS; ++} ++ + static void v1_disk2mem_dqblk(struct mem_dqblk *m, struct v1_disk_dqblk *d) + { + m->dqb_ihardlimit = d->dqb_ihardlimit; + m->dqb_isoftlimit = d->dqb_isoftlimit; + m->dqb_curinodes = d->dqb_curinodes; +- m->dqb_bhardlimit = d->dqb_bhardlimit; +- m->dqb_bsoftlimit = d->dqb_bsoftlimit; +- m->dqb_curspace = ((qsize_t)d->dqb_curblocks) << QUOTABLOCK_BITS; ++ m->dqb_bhardlimit = v1_qbtos(d->dqb_bhardlimit); ++ m->dqb_bsoftlimit = v1_qbtos(d->dqb_bsoftlimit); ++ m->dqb_curspace = v1_qbtos(d->dqb_curblocks); + m->dqb_itime = d->dqb_itime; + m->dqb_btime = d->dqb_btime; + } +@@ -31,9 +44,9 @@ static void v1_mem2disk_dqblk(struct v1_disk_dqblk *d, struct mem_dqblk *m) + d->dqb_ihardlimit = m->dqb_ihardlimit; + d->dqb_isoftlimit = m->dqb_isoftlimit; + d->dqb_curinodes = m->dqb_curinodes; +- d->dqb_bhardlimit = m->dqb_bhardlimit; +- d->dqb_bsoftlimit = m->dqb_bsoftlimit; +- d->dqb_curblocks = toqb(m->dqb_curspace); ++ d->dqb_bhardlimit = v1_stoqb(m->dqb_bhardlimit); ++ d->dqb_bsoftlimit = v1_stoqb(m->dqb_bsoftlimit); ++ d->dqb_curblocks = v1_stoqb(m->dqb_curspace); + d->dqb_itime = m->dqb_itime; + d->dqb_btime = m->dqb_btime; + } +diff --git a/fs/quota_v2.c b/fs/quota_v2.c +index b53827d..51c4717 100644 +--- a/fs/quota_v2.c ++++ b/fs/quota_v2.c +@@ -26,6 +26,19 @@ typedef char *dqbuf_t; + #define GETIDINDEX(id, depth) (((id) >> ((V2_DQTREEDEPTH-(depth)-1)*8)) & 0xff) + #define GETENTRIES(buf) ((struct v2_disk_dqblk *)(((char *)buf)+sizeof(struct v2_disk_dqdbheader))) + ++#define QUOTABLOCK_BITS 10 ++#define QUOTABLOCK_SIZE (1 << QUOTABLOCK_BITS) ++ ++static inline qsize_t v2_stoqb(qsize_t space) ++{ ++ return (space + QUOTABLOCK_SIZE - 1) >> QUOTABLOCK_BITS; ++} ++ ++static inline qsize_t v2_qbtos(qsize_t blocks) ++{ ++ return blocks << QUOTABLOCK_BITS; ++} ++ + /* Check whether given file is really vfsv0 quotafile */ + static int v2_check_quota_file(struct super_block *sb, int type) + { +@@ -104,8 +117,8 @@ static void disk2memdqb(struct mem_dqblk *m, struct v2_disk_dqblk *d) + m->dqb_isoftlimit = le32_to_cpu(d->dqb_isoftlimit); + m->dqb_curinodes = le32_to_cpu(d->dqb_curinodes); + m->dqb_itime = le64_to_cpu(d->dqb_itime); +- m->dqb_bhardlimit = le32_to_cpu(d->dqb_bhardlimit); +- m->dqb_bsoftlimit = le32_to_cpu(d->dqb_bsoftlimit); ++ m->dqb_bhardlimit = v2_qbtos(le32_to_cpu(d->dqb_bhardlimit)); ++ m->dqb_bsoftlimit = v2_qbtos(le32_to_cpu(d->dqb_bsoftlimit)); + m->dqb_curspace = le64_to_cpu(d->dqb_curspace); + m->dqb_btime = le64_to_cpu(d->dqb_btime); + } +@@ -116,8 +129,8 @@ static void mem2diskdqb(struct v2_disk_dqblk *d, struct mem_dqblk *m, qid_t id) + d->dqb_isoftlimit = cpu_to_le32(m->dqb_isoftlimit); + d->dqb_curinodes = cpu_to_le32(m->dqb_curinodes); + d->dqb_itime = cpu_to_le64(m->dqb_itime); +- d->dqb_bhardlimit = cpu_to_le32(m->dqb_bhardlimit); +- d->dqb_bsoftlimit = cpu_to_le32(m->dqb_bsoftlimit); ++ d->dqb_bhardlimit = cpu_to_le32(v2_qbtos(m->dqb_bhardlimit)); ++ d->dqb_bsoftlimit = cpu_to_le32(v2_qbtos(m->dqb_bsoftlimit)); + d->dqb_curspace = cpu_to_le64(m->dqb_curspace); + d->dqb_btime = cpu_to_le64(m->dqb_btime); + d->dqb_id = cpu_to_le32(id); +diff --git a/include/linux/quota.h b/include/linux/quota.h +index eeae7a9..5167786 100644 +--- a/include/linux/quota.h ++++ b/include/linux/quota.h +@@ -41,15 +41,6 @@ + #define __DQUOT_VERSION__ "dquot_6.5.1" + #define __DQUOT_NUM_VERSION__ 6*10000+5*100+1 + +-/* Size of blocks in which are counted size limits */ +-#define QUOTABLOCK_BITS 10 +-#define QUOTABLOCK_SIZE (1 << QUOTABLOCK_BITS) +- +-/* Conversion routines from and to quota blocks */ +-#define qb2kb(x) ((x) << (QUOTABLOCK_BITS-10)) +-#define kb2qb(x) ((x) >> (QUOTABLOCK_BITS-10)) +-#define toqb(x) (((x) + QUOTABLOCK_SIZE - 1) >> QUOTABLOCK_BITS) +- + #define MAXQUOTAS 2 + #define USRQUOTA 0 /* element used for user quotas */ + #define GRPQUOTA 1 /* element used for group quotas */ +@@ -82,6 +73,11 @@ + #define Q_GETQUOTA 0x800007 /* get user quota structure */ + #define Q_SETQUOTA 0x800008 /* set user quota structure */ + ++/* Size of block in which space limits are passed through the quota ++ * interface */ ++#define QIF_DQBLKSIZE_BITS 10 ++#define QIF_DQBLKSIZE (1 << QIF_DQBLKSIZE_BITS) ++ + /* + * Quota structure used for communication with userspace via quotactl + * Following flags are used to specify which fields are valid +@@ -189,12 +185,12 @@ extern spinlock_t dq_data_lock; + * Data for one user/group kept in memory + */ + struct mem_dqblk { +- __u32 dqb_bhardlimit; /* absolute limit on disk blks alloc */ +- __u32 dqb_bsoftlimit; /* preferred limit on disk blks */ ++ qsize_t dqb_bhardlimit; /* absolute limit on disk blks alloc */ ++ qsize_t dqb_bsoftlimit; /* preferred limit on disk blks */ + qsize_t dqb_curspace; /* current used space */ +- __u32 dqb_ihardlimit; /* absolute limit on allocated inodes */ +- __u32 dqb_isoftlimit; /* preferred inode limit */ +- __u32 dqb_curinodes; /* current # allocated inodes */ ++ qsize_t dqb_ihardlimit; /* absolute limit on allocated inodes */ ++ qsize_t dqb_isoftlimit; /* preferred inode limit */ ++ qsize_t dqb_curinodes; /* current # allocated inodes */ + time_t dqb_btime; /* time limit for excessive disk use */ + time_t dqb_itime; /* time limit for excessive inode use */ + }; +@@ -289,9 +285,9 @@ struct dquot_operations { + int (*initialize) (struct inode *, int); + int (*drop) (struct inode *); + int (*alloc_space) (struct inode *, qsize_t, int); +- int (*alloc_inode) (const struct inode *, unsigned long); ++ int (*alloc_inode) (const struct inode *, qsize_t); + int (*free_space) (struct inode *, qsize_t); +- int (*free_inode) (const struct inode *, unsigned long); ++ int (*free_inode) (const struct inode *, qsize_t); + int (*transfer) (struct inode *, struct iattr *); + int (*write_dquot) (struct dquot *); /* Ordinary dquot write */ + struct dquot *(*alloc_dquot)(struct super_block *, int); /* Allocate memory for new dquot (can be NULL if no special entries dquot are needed) */ +diff --git a/include/linux/quotaops.h b/include/linux/quotaops.h +index ca6b9b5..9e7bc4b 100644 +--- a/include/linux/quotaops.h ++++ b/include/linux/quotaops.h +@@ -29,10 +29,10 @@ int dquot_initialize(struct inode *inode, int type); + int dquot_drop(struct inode *inode); + + int dquot_alloc_space(struct inode *inode, qsize_t number, int prealloc); +-int dquot_alloc_inode(const struct inode *inode, unsigned long number); ++int dquot_alloc_inode(const struct inode *inode, qsize_t number); + + int dquot_free_space(struct inode *inode, qsize_t number); +-int dquot_free_inode(const struct inode *inode, unsigned long number); ++int dquot_free_inode(const struct inode *inode, qsize_t number); + + int dquot_transfer(struct inode *inode, struct iattr *iattr); + int dquot_commit(struct dquot *dquot); +-- +1.5.2.4 + --- lustre-1.8.3.orig/lustre/kernel_patches/patches/jbd2-jcberr-2.6.26-vanilla.patch +++ lustre-1.8.3/lustre/kernel_patches/patches/jbd2-jcberr-2.6.26-vanilla.patch @@ -0,0 +1,211 @@ +Index: linux-source-2.6.26/include/linux/jbd2.h +=================================================================== +--- linux-source-2.6.26.orig/include/linux/jbd2.h 2008-07-13 23:51:29.000000000 +0200 ++++ linux-source-2.6.26/include/linux/jbd2.h 2009-10-09 13:42:21.000000000 +0200 +@@ -379,6 +379,27 @@ + bit_spin_unlock(BH_JournalHead, &bh->b_state); + } + ++#define HAVE_JOURNAL_CALLBACK_STATUS ++/** ++ * struct journal_callback - Base structure for callback information. ++ * @jcb_list: list information for other callbacks attached to the same handle. ++ * @jcb_func: Function to call with this callback structure. ++ * ++ * This struct is a 'seed' structure for a using with your own callback ++ * structs. If you are using callbacks you must allocate one of these ++ * or another struct of your own definition which has this struct ++ * as it's first element and pass it to journal_callback_set(). ++ * ++ * This is used internally by jbd2 to maintain callback information. ++ * ++ * See journal_callback_set for more information. ++ **/ ++struct journal_callback { ++ struct list_head jcb_list; /* t_jcb_lock */ ++ void (*jcb_func)(struct journal_callback *jcb, int error); ++ /* user data goes here */ ++}; ++ + struct jbd2_revoke_table_s; + + /** +@@ -387,6 +408,7 @@ + * @h_transaction: Which compound transaction is this update a part of? + * @h_buffer_credits: Number of remaining buffers we are allowed to dirty. + * @h_ref: Reference count on this handle ++ * @h_jcb: List of application registered callbacks for this handle. + * @h_err: Field for caller's use to track errors through large fs operations + * @h_sync: flag for sync-on-close + * @h_jdata: flag to force data journaling +@@ -412,6 +434,13 @@ + /* operations */ + int h_err; + ++ /* ++ * List of application registered callbacks for this handle. The ++ * function(s) will be called after the transaction that this handle is ++ * part of has been committed to disk. [t_jcb_lock] ++ */ ++ struct list_head h_jcb; ++ + /* Flags [no locking] */ + unsigned int h_sync: 1; /* sync-on-close */ + unsigned int h_jdata: 1; /* force data journaling */ +@@ -467,6 +496,8 @@ + * j_state_lock + * ->j_list_lock (journal_unmap_buffer) + * ++ * t_handle_lock ++ * ->t_jcb_lock + */ + + struct transaction_s +@@ -613,6 +644,15 @@ + */ + int t_handle_count; + ++ /* ++ * Protects the callback list ++ */ ++ spinlock_t t_jcb_lock; ++ /* ++ * List of registered callback functions for this transaction. ++ * Called when the transaction is committed. [t_jcb_lock] ++ */ ++ struct list_head t_jcb; + }; + + struct transaction_run_stats_s { +@@ -1016,6 +1056,9 @@ + extern int jbd2_journal_flush (journal_t *); + extern void jbd2_journal_lock_updates (journal_t *); + extern void jbd2_journal_unlock_updates (journal_t *); ++extern void jbd2_journal_callback_set(handle_t *handle, ++ void (*fn)(struct journal_callback *,int), ++ struct journal_callback *jcb); + + extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, + struct block_device *fs_dev, +Index: linux-source-2.6.26/fs/jbd2/checkpoint.c +=================================================================== +--- linux-source-2.6.26.orig/fs/jbd2/checkpoint.c 2008-07-13 23:51:29.000000000 +0200 ++++ linux-source-2.6.26/fs/jbd2/checkpoint.c 2009-10-09 13:42:21.000000000 +0200 +@@ -696,6 +696,7 @@ + J_ASSERT(transaction->t_checkpoint_list == NULL); + J_ASSERT(transaction->t_checkpoint_io_list == NULL); + J_ASSERT(transaction->t_updates == 0); ++ J_ASSERT(list_empty(&transaction->t_jcb)); + J_ASSERT(journal->j_committing_transaction != transaction); + J_ASSERT(journal->j_running_transaction != transaction); + +Index: linux-source-2.6.26/fs/jbd2/commit.c +=================================================================== +--- linux-source-2.6.26.orig/fs/jbd2/commit.c 2008-07-13 23:51:29.000000000 +0200 ++++ linux-source-2.6.26/fs/jbd2/commit.c 2009-10-09 13:42:21.000000000 +0200 +@@ -854,6 +854,30 @@ + /* AKPM: bforget here */ + } + ++ /* ++ * Call any callbacks that had been registered for handles in this ++ * transaction. It is up to the callback to free any allocated ++ * memory. ++ * ++ * The spinlocking (t_jcb_lock) here is surely unnecessary... ++ */ ++ spin_lock(&commit_transaction->t_jcb_lock); ++ if (!list_empty(&commit_transaction->t_jcb)) { ++ struct list_head *p, *n; ++ int error = is_journal_aborted(journal); ++ ++ list_for_each_safe(p, n, &commit_transaction->t_jcb) { ++ struct journal_callback *jcb; ++ ++ jcb = list_entry(p, struct journal_callback, jcb_list); ++ list_del(p); ++ spin_unlock(&commit_transaction->t_jcb_lock); ++ jcb->jcb_func(jcb, error); ++ spin_lock(&commit_transaction->t_jcb_lock); ++ } ++ } ++ spin_unlock(&commit_transaction->t_jcb_lock); ++ + jbd_debug(3, "JBD: commit phase 6\n"); + + if (!JBD2_HAS_INCOMPAT_FEATURE(journal, +Index: linux-source-2.6.26/fs/jbd2/journal.c +=================================================================== +--- linux-source-2.6.26.orig/fs/jbd2/journal.c 2009-10-09 13:39:04.000000000 +0200 ++++ linux-source-2.6.26/fs/jbd2/journal.c 2009-10-09 13:43:03.000000000 +0200 +@@ -82,6 +82,9 @@ + EXPORT_SYMBOL(jbd2_journal_invalidatepage); + EXPORT_SYMBOL(jbd2_journal_try_to_free_buffers); + EXPORT_SYMBOL(jbd2_journal_force_commit); ++EXPORT_SYMBOL(jbd2_journal_callback_set); ++EXPORT_SYMBOL(jbd2_journal_bmap); ++ + + static int journal_convert_superblock_v1(journal_t *, journal_superblock_t *); + static void __journal_abort_soft (journal_t *journal, int errno); +Index: linux-source-2.6.26/fs/jbd2/transaction.c +=================================================================== +--- linux-source-2.6.26.orig/fs/jbd2/transaction.c 2008-07-13 23:51:29.000000000 +0200 ++++ linux-source-2.6.26/fs/jbd2/transaction.c 2009-10-09 13:42:21.000000000 +0200 +@@ -252,6 +252,7 @@ + memset(handle, 0, sizeof(*handle)); + handle->h_buffer_credits = nblocks; + handle->h_ref = 1; ++ INIT_LIST_HEAD(&handle->h_jcb); + + lockdep_init_map(&handle->h_lockdep_map, "jbd2_handle", + &jbd2_handle_key, 0); +@@ -1350,6 +1351,36 @@ + } + + /** ++ * void jbd2_journal_callback_set() - Register a callback function for this handle. ++ * @handle: handle to attach the callback to. ++ * @func: function to callback. ++ * @jcb: structure with additional information required by func() , and ++ * some space for jbd2 internal information. ++ * ++ * The function will be ++ * called when the transaction that this handle is part of has been ++ * committed to disk with the original callback data struct and the ++ * error status of the journal as parameters. There is no guarantee of ++ * ordering between handles within a single transaction, nor between ++ * callbacks registered on the same handle. ++ * ++ * The caller is responsible for allocating the journal_callback struct. ++ * This is to allow the caller to add as much extra data to the callback ++ * as needed, but reduce the overhead of multiple allocations. The caller ++ * allocated struct must start with a struct journal_callback at offset 0, ++ * and has the caller-specific data afterwards. ++ */ ++void jbd2_journal_callback_set(handle_t *handle, ++ void (*func)(struct journal_callback *jcb, int error), ++ struct journal_callback *jcb) ++{ ++ spin_lock(&handle->h_transaction->t_jcb_lock); ++ list_add_tail(&jcb->jcb_list, &handle->h_jcb); ++ spin_unlock(&handle->h_transaction->t_jcb_lock); ++ jcb->jcb_func = func; ++} ++ ++/** + * int jbd2_journal_stop() - complete a transaction + * @handle: tranaction to complete. + * +@@ -1423,6 +1454,11 @@ + wake_up(&journal->j_wait_transaction_locked); + } + ++ /* Move callbacks from the handle to the transaction. */ ++ spin_lock(&transaction->t_jcb_lock); ++ list_splice(&handle->h_jcb, &transaction->t_jcb); ++ spin_unlock(&transaction->t_jcb_lock); ++ + /* + * If the handle is marked SYNC, we need to set another commit + * going! We also want to force a commit if the current --- lustre-1.8.3.orig/lustre/kernel_patches/series/2.6.26-vanilla.series +++ lustre-1.8.3/lustre/kernel_patches/series/2.6.26-vanilla.series @@ -0,0 +1,2 @@ +lustre_version.patch +debian-2.6.26.diff --- lustre-1.8.3.orig/lustre/kernel_patches/series/2.6.27-vanilla.series +++ lustre-1.8.3/lustre/kernel_patches/series/2.6.27-vanilla.series @@ -0,0 +1,15 @@ +lustre_version.patch +vfs_races-2.6.22-vanilla.patch +iopen-misc-2.6.22-vanilla.patch +export-truncate-2.6.18-vanilla.patch +export_symbols-2.6.22-vanilla.patch +dev_read_only-2.6.27-vanilla.patch +export-2.6.27-vanilla.patch +export-show_task-2.6.27-vanilla.patch +sd_iostats-2.6.27-vanilla.patch +md-mmp-unplug-dev-2.6.27-vanilla.patch +quota-Split-off-quota-tree-handling-into-a-separate.patch +quota-Increase-size-of-variables-for-limits-and-ino.patch +quota-support-64-bit-quota-format-2.6.27-vanilla.patch +jbd2-jcberr-2.6-sles11.patch +jbd2-commit-timer-no-jiffies-rounding.diff