X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;ds=sidebyside;f=lustre%2Fllite%2Flproc_llite.c;h=03dd0acc83bbea3bca113e487af9dbd33f75b58d;hb=5900da4464d90468519964571c441382cd512232;hp=5f429c52bbbd8a2f46e4e909944d66c9874180f1;hpb=1aa62294d88e234f54cc2de573480591739b8754;p=fs%2Flustre-release.git diff --git a/lustre/llite/lproc_llite.c b/lustre/llite/lproc_llite.c index 5f429c5..03dd0ac 100644 --- a/lustre/llite/lproc_llite.c +++ b/lustre/llite/lproc_llite.c @@ -22,32 +22,22 @@ #define DEBUG_SUBSYSTEM S_LLITE #include -#include -#include +#include +#include #include -#include -#ifdef HAVE_MM_INLINE -#include -#endif +#include #include "llite_internal.h" -/* /proc/lustre/llite mount point registration */ struct proc_dir_entry *proc_lustre_fs_root; + +#ifdef LPROCFS +/* /proc/lustre/llite mount point registration */ struct file_operations llite_dump_pgcache_fops; struct file_operations ll_ra_stats_fops; -struct file_operations llite_wait_times_fops; - -#ifndef LPROCFS -int lprocfs_register_mountpoint(struct proc_dir_entry *parent, - struct super_block *sb, char *osc, char *mdc) -{ - return 0; -} -void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi){} -#else - -long long mnt_instance; +struct file_operations ll_rw_extents_stats_fops; +struct file_operations ll_rw_extents_stats_pp_fops; +struct file_operations ll_rw_offset_stats_fops; static int ll_rd_blksize(char *page, char **start, off_t off, int count, int *eof, void *data) @@ -57,7 +47,7 @@ static int ll_rd_blksize(char *page, char **start, off_t off, int count, int rc; LASSERT(sb != NULL); - rc = ll_statfs_internal(sb, &osfs, jiffies - HZ); + rc = ll_statfs_internal(sb, &osfs, cfs_time_current_64() - HZ); if (!rc) { *eof = 1; rc = snprintf(page, count, "%u\n", osfs.os_bsize); @@ -74,7 +64,7 @@ static int ll_rd_kbytestotal(char *page, char **start, off_t off, int count, int rc; LASSERT(sb != NULL); - rc = ll_statfs_internal(sb, &osfs, jiffies - HZ); + rc = ll_statfs_internal(sb, &osfs, cfs_time_current_64() - HZ); if (!rc) { __u32 blk_size = osfs.os_bsize >> 10; __u64 result = osfs.os_blocks; @@ -97,7 +87,7 @@ static int ll_rd_kbytesfree(char *page, char **start, off_t off, int count, int rc; LASSERT(sb != NULL); - rc = ll_statfs_internal(sb, &osfs, jiffies - HZ); + rc = ll_statfs_internal(sb, &osfs, cfs_time_current_64() - HZ); if (!rc) { __u32 blk_size = osfs.os_bsize >> 10; __u64 result = osfs.os_bfree; @@ -119,7 +109,7 @@ static int ll_rd_kbytesavail(char *page, char **start, off_t off, int count, int rc; LASSERT(sb != NULL); - rc = ll_statfs_internal(sb, &osfs, jiffies - HZ); + rc = ll_statfs_internal(sb, &osfs, cfs_time_current_64() - HZ); if (!rc) { __u32 blk_size = osfs.os_bsize >> 10; __u64 result = osfs.os_bavail; @@ -141,7 +131,7 @@ static int ll_rd_filestotal(char *page, char **start, off_t off, int count, int rc; LASSERT(sb != NULL); - rc = ll_statfs_internal(sb, &osfs, jiffies - HZ); + rc = ll_statfs_internal(sb, &osfs, cfs_time_current_64() - HZ); if (!rc) { *eof = 1; rc = snprintf(page, count, LPU64"\n", osfs.os_files); @@ -157,7 +147,7 @@ static int ll_rd_filesfree(char *page, char **start, off_t off, int count, int rc; LASSERT(sb != NULL); - rc = ll_statfs_internal(sb, &osfs, jiffies - HZ); + rc = ll_statfs_internal(sb, &osfs, cfs_time_current_64() - HZ); if (!rc) { *eof = 1; rc = snprintf(page, count, LPU64"\n", osfs.os_ffree); @@ -166,6 +156,23 @@ static int ll_rd_filesfree(char *page, char **start, off_t off, int count, } +static int ll_rd_client_type(char *page, char **start, off_t off, int count, + int *eof, void *data) +{ + struct ll_sb_info *sbi = ll_s2sbi((struct super_block *)data); + int rc; + + LASSERT(sbi != NULL); + + *eof = 1; + if (sbi->ll_flags & LL_SBI_RMT_CLIENT) + rc = snprintf(page, count, "remote client\n"); + else + rc = snprintf(page, count, "local client\n"); + + return rc; +} + static int ll_rd_fstype(char *page, char **start, off_t off, int count, int *eof, void *data) { @@ -186,260 +193,267 @@ static int ll_rd_sb_uuid(char *page, char **start, off_t off, int count, return snprintf(page, count, "%s\n", ll_s2sbi(sb)->ll_sb_uuid.uuid); } -static int ll_rd_read_ahead(char *page, char **start, off_t off, int count, - int *eof, void *data) +static int ll_rd_max_readahead_mb(char *page, char **start, off_t off, + int count, int *eof, void *data) { - struct super_block *sb = (struct super_block*)data; + struct super_block *sb = data; struct ll_sb_info *sbi = ll_s2sbi(sb); - int val, rc; - ENTRY; + long pages_number; + int mult; - *eof = 1; - val = (sbi->ll_flags & LL_SBI_READAHEAD) ? 1 : 0; - rc = snprintf(page, count, "%d\n", val); - RETURN(rc); + spin_lock(&sbi->ll_lock); + pages_number = sbi->ll_ra_info.ra_max_pages; + spin_unlock(&sbi->ll_lock); + + mult = 1 << (20 - PAGE_CACHE_SHIFT); + return lprocfs_read_frac_helper(page, count, pages_number, mult); } -static int ll_wr_read_ahead(struct file *file, const char *buffer, - unsigned long count, void *data) +static int ll_wr_max_readahead_mb(struct file *file, const char *buffer, + unsigned long count, void *data) { - struct super_block *sb = (struct super_block*)data; + struct super_block *sb = data; struct ll_sb_info *sbi = ll_s2sbi(sb); - int readahead; - ENTRY; - - if (sscanf(buffer, "%d", &readahead) != 1) - RETURN(-EINVAL); + int mult, rc, pages_number; - if (readahead) - sbi->ll_flags |= LL_SBI_READAHEAD; - else - sbi->ll_flags &= ~LL_SBI_READAHEAD; + mult = 1 << (20 - CFS_PAGE_SHIFT); + rc = lprocfs_write_frac_helper(buffer, count, &pages_number, mult); + if (rc) + return rc; - RETURN(count); -} + if (pages_number < 0 || pages_number > num_physpages / 2) { + CERROR("can't set file readahead more than %lu MB\n", + num_physpages >> (20 - CFS_PAGE_SHIFT + 1)); /*1/2 of RAM*/ + return -ERANGE; + } -static int ll_wr_config_update(struct file *file, const char *buffer, - unsigned long count, void *data) -{ - struct super_block *sb = (struct super_block*)data; - struct ll_sb_info *sbi = ll_s2sbi(sb); - ENTRY; + spin_lock(&sbi->ll_lock); + sbi->ll_ra_info.ra_max_pages = pages_number; + spin_unlock(&sbi->ll_lock); - CWARN("Starting a LOV/OST update !\n"); - ll_process_config_update(sbi, 0); - RETURN(count); + return count; } -static int ll_rd_max_read_ahead_mb(char *page, char **start, off_t off, - int count, int *eof, void *data) +static int ll_rd_max_read_ahead_whole_mb(char *page, char **start, off_t off, + int count, int *eof, void *data) { struct super_block *sb = data; struct ll_sb_info *sbi = ll_s2sbi(sb); - unsigned val; + long pages_number; + int mult; spin_lock(&sbi->ll_lock); - val = (sbi->ll_ra_info.ra_max_pages << PAGE_CACHE_SHIFT) >> 20; + pages_number = sbi->ll_ra_info.ra_max_read_ahead_whole_pages; spin_unlock(&sbi->ll_lock); - return snprintf(page, count, "%u\n", val); + mult = 1 << (20 - CFS_PAGE_SHIFT); + return lprocfs_read_frac_helper(page, count, pages_number, mult); } -static int ll_wr_max_read_ahead_mb(struct file *file, const char *buffer, - unsigned long count, void *data) +static int ll_wr_max_read_ahead_whole_mb(struct file *file, const char *buffer, + unsigned long count, void *data) { struct super_block *sb = data; struct ll_sb_info *sbi = ll_s2sbi(sb); - int val, rc; + int mult, rc, pages_number; - rc = lprocfs_write_helper(buffer, count, &val); + mult = 1 << (20 - CFS_PAGE_SHIFT); + rc = lprocfs_write_frac_helper(buffer, count, &pages_number, mult); if (rc) return rc; - if (val < 0 || val > (num_physpages << PAGE_SHIFT) >> 20) + /* Cap this at the current max readahead window size, the readahead + * algorithm does this anyway so it's pointless to set it larger. */ + if (pages_number < 0 || pages_number > sbi->ll_ra_info.ra_max_pages) { + CERROR("can't set max_read_ahead_whole_mb more than " + "max_read_ahead_mb: %lu\n", + sbi->ll_ra_info.ra_max_pages >> (20 - CFS_PAGE_SHIFT)); return -ERANGE; + } spin_lock(&sbi->ll_lock); - sbi->ll_ra_info.ra_max_pages = (val << 20) >> PAGE_CACHE_SHIFT; + sbi->ll_ra_info.ra_max_read_ahead_whole_pages = pages_number; spin_unlock(&sbi->ll_lock); return count; } -static int ll_rd_gns_upcall(char *page, char **start, off_t off, - int count, int *eof, void *data) +static int ll_rd_max_cached_mb(char *page, char **start, off_t off, + int count, int *eof, void *data) { - struct super_block *sb = (struct super_block *)data; + struct super_block *sb = data; struct ll_sb_info *sbi = ll_s2sbi(sb); - int len; + long pages_number; + int mult; - down(&sbi->ll_gns_sem); - len = snprintf(page, count, "%s\n", sbi->ll_gns_upcall); - up(&sbi->ll_gns_sem); + spin_lock(&sbi->ll_lock); + pages_number = sbi->ll_async_page_max; + spin_unlock(&sbi->ll_lock); - return len; + mult = 1 << (20 - CFS_PAGE_SHIFT); + return lprocfs_read_frac_helper(page, count, pages_number, mult);; } -static int ll_wr_gns_upcall(struct file *file, const char *buffer, - unsigned long count, void *data) +static int ll_wr_max_cached_mb(struct file *file, const char *buffer, + unsigned long count, void *data) { - struct super_block *sb = (struct super_block *)data; + struct super_block *sb = data; struct ll_sb_info *sbi = ll_s2sbi(sb); + int mult, rc, pages_number; - down(&sbi->ll_gns_sem); - - /* - * upcall should not be the same as object name, check for possible - * overflow. - */ - if (count < sizeof(sbi->ll_gns_upcall) && - (strlen(sbi->ll_gns_oname) != count || - strncmp(sbi->ll_gns_oname, buffer, count))) - snprintf(sbi->ll_gns_upcall, count, "%s", buffer); - - up(&sbi->ll_gns_sem); - return count; -} - -static int ll_rd_gns_object_name(char *page, char **start, off_t off, - int count, int *eof, void *data) -{ - struct super_block *sb = (struct super_block *)data; - struct ll_sb_info *sbi = ll_s2sbi(sb); - int len; + mult = 1 << (20 - CFS_PAGE_SHIFT); + rc = lprocfs_write_frac_helper(buffer, count, &pages_number, mult); + if (rc) + return rc; - down(&sbi->ll_gns_sem); - len = snprintf(page, count, "%s\n", sbi->ll_gns_oname); - up(&sbi->ll_gns_sem); + if (pages_number < 0 || pages_number > num_physpages) { + CERROR("can't set max cache more than %lu MB\n", + num_physpages >> (20 - CFS_PAGE_SHIFT)); + return -ERANGE; + } - return len; -} + spin_lock(&sbi->ll_lock); + sbi->ll_async_page_max = pages_number ; + spin_unlock(&sbi->ll_lock); + + if (!sbi->ll_dt_exp) + /* Not set up yet, don't call llap_shrink_cache */ + return count; -static int ll_wr_gns_object_name(struct file *file, const char *buffer, - unsigned long count, void *data) -{ - struct super_block *sb = (struct super_block *)data; - struct ll_sb_info *sbi = ll_s2sbi(sb); + if (sbi->ll_async_page_count >= sbi->ll_async_page_max) + llap_shrink_cache(sbi, 0); - /* checking for setting "." and ".." as object name */ - if (buffer[0] == '.') switch (count) { - case 2: - /* this is "." case with terminating zero */ - return -EINVAL; - case 3: - /* this is ".." case with terminating zero */ - if (buffer[1] == '.') - return -EINVAL; - } - - down(&sbi->ll_gns_sem); - - /* - * upcall should not be the same as object name, check for possible - * overflow. - */ - if (count < sizeof(sbi->ll_gns_oname) && - (strlen(sbi->ll_gns_upcall) != count || - strncmp(sbi->ll_gns_upcall, buffer, count))) - snprintf(sbi->ll_gns_oname, count, "%s", buffer); - - up(&sbi->ll_gns_sem); return count; } -static int ll_rd_gns_timeout(char *page, char **start, off_t off, - int count, int *eof, void *data) +static int ll_rd_checksum(char *page, char **start, off_t off, + int count, int *eof, void *data) { - struct super_block *sb = (struct super_block *)data; + struct super_block *sb = data; struct ll_sb_info *sbi = ll_s2sbi(sb); - int len; - - down(&sbi->ll_gns_sem); - len = snprintf(page, count, "%lu\n", - (unsigned long)sbi->ll_gns_timeout); - up(&sbi->ll_gns_sem); - return len; + return snprintf(page, count, "%u\n", + (sbi->ll_flags & LL_SBI_CHECKSUM) ? 1 : 0); } -static int ll_wr_gns_timeout(struct file *file, const char *buffer, - unsigned long count, void *data) +static int ll_wr_checksum(struct file *file, const char *buffer, + unsigned long count, void *data) { - struct super_block *sb = (struct super_block *)data; + struct super_block *sb = data; struct ll_sb_info *sbi = ll_s2sbi(sb); int val, rc; + if (!sbi->ll_dt_exp) + /* Not set up yet */ + return -EAGAIN; + rc = lprocfs_write_helper(buffer, count, &val); if (rc) return rc; + if (val) + sbi->ll_flags |= LL_SBI_CHECKSUM; + else + sbi->ll_flags &= ~LL_SBI_CHECKSUM; - down(&sbi->ll_gns_sem); - if (val > sbi->ll_gns_tick) - sbi->ll_gns_timeout = val; - up(&sbi->ll_gns_sem); + rc = obd_set_info_async(sbi->ll_dt_exp, strlen("checksum"), "checksum", + sizeof(val), &val, NULL); + if (rc) + CWARN("Failed to set OSC checksum flags: %d\n", rc); return count; } -static int ll_rd_gns_tick(char *page, char **start, off_t off, +static int ll_rd_max_rw_chunk(char *page, char **start, off_t off, int count, int *eof, void *data) { - struct super_block *sb = (struct super_block *)data; - struct ll_sb_info *sbi = ll_s2sbi(sb); - int len; - - down(&sbi->ll_gns_sem); - len = snprintf(page, count, "%lu\n", - (unsigned long)sbi->ll_gns_tick); - up(&sbi->ll_gns_sem); + struct super_block *sb = data; - return len; + return snprintf(page, count, "%lu\n", ll_s2sbi(sb)->ll_max_rw_chunk); } -static int ll_wr_gns_tick(struct file *file, const char *buffer, +static int ll_wr_max_rw_chunk(struct file *file, const char *buffer, unsigned long count, void *data) { - struct super_block *sb = (struct super_block *)data; - struct ll_sb_info *sbi = ll_s2sbi(sb); - int val, rc; + struct super_block *sb = data; + int rc, val; rc = lprocfs_write_helper(buffer, count, &val); if (rc) return rc; - - down(&sbi->ll_gns_sem); - if (val < sbi->ll_gns_timeout) - sbi->ll_gns_tick = val; - up(&sbi->ll_gns_sem); - + ll_s2sbi(sb)->ll_max_rw_chunk = val; return count; } -static int ll_rd_gns_enabled(char *page, char **start, off_t off, - int count, int *eof, void *data) +static int ll_rd_track_id(char *page, int count, void *data, + enum stats_track_type type) { - struct super_block *sb = (struct super_block *)data; - struct ll_sb_info *sbi = ll_s2sbi(sb); + struct super_block *sb = data; + + if (ll_s2sbi(sb)->ll_stats_track_type == type) { + return snprintf(page, count, "%d\n", + ll_s2sbi(sb)->ll_stats_track_id); - return snprintf(page, count, "%d\n", - atomic_read(&sbi->ll_gns_enabled)); + } else if (ll_s2sbi(sb)->ll_stats_track_type == STATS_TRACK_ALL) { + return snprintf(page, count, "0 (all)\n"); + } else { + return snprintf(page, count, "untracked\n"); + } } -static int ll_wr_gns_enabled(struct file *file, const char *buffer, - unsigned long count, void *data) +static int ll_wr_track_id(const char *buffer, unsigned long count, void *data, + enum stats_track_type type) { - struct super_block *sb = (struct super_block *)data; - struct ll_sb_info *sbi = ll_s2sbi(sb); - int val, rc; + struct super_block *sb = data; + int rc, pid; - rc = lprocfs_write_helper(buffer, count, &val); + rc = lprocfs_write_helper(buffer, count, &pid); if (rc) return rc; - - atomic_set(&sbi->ll_gns_enabled, (val != 0 ? 1 : 0)); + ll_s2sbi(sb)->ll_stats_track_id = pid; + if (pid == 0) + ll_s2sbi(sb)->ll_stats_track_type = STATS_TRACK_ALL; + else + ll_s2sbi(sb)->ll_stats_track_type = type; + lprocfs_clear_stats(ll_s2sbi(sb)->ll_stats); return count; } +static int ll_rd_track_pid(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + return (ll_rd_track_id(page, count, data, STATS_TRACK_PID)); +} + +static int ll_wr_track_pid(struct file *file, const char *buffer, + unsigned long count, void *data) +{ + return (ll_wr_track_id(buffer, count, data, STATS_TRACK_PID)); +} + +static int ll_rd_track_ppid(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + return (ll_rd_track_id(page, count, data, STATS_TRACK_PPID)); +} + +static int ll_wr_track_ppid(struct file *file, const char *buffer, + unsigned long count, void *data) +{ + return (ll_wr_track_id(buffer, count, data, STATS_TRACK_PPID)); +} + +static int ll_rd_track_gid(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + return (ll_rd_track_id(page, count, data, STATS_TRACK_GID)); +} + +static int ll_wr_track_gid(struct file *file, const char *buffer, + unsigned long count, void *data) +{ + return (ll_wr_track_id(buffer, count, data, STATS_TRACK_GID)); +} + static struct lprocfs_vars lprocfs_obd_vars[] = { { "uuid", ll_rd_sb_uuid, 0, 0 }, //{ "mntpt_path", ll_rd_path, 0, 0 }, @@ -450,27 +464,18 @@ static struct lprocfs_vars lprocfs_obd_vars[] = { { "kbytesavail", ll_rd_kbytesavail, 0, 0 }, { "filestotal", ll_rd_filestotal, 0, 0 }, { "filesfree", ll_rd_filesfree, 0, 0 }, + { "client_type", ll_rd_client_type, 0, 0 }, //{ "filegroups", lprocfs_rd_filegroups, 0, 0 }, - { "read_ahead", ll_rd_read_ahead, ll_wr_read_ahead, 0 }, - { "config_update", 0, ll_wr_config_update, 0 }, - { "max_read_ahead_mb", ll_rd_max_read_ahead_mb, - ll_wr_max_read_ahead_mb, 0 }, - - { "gns_upcall", ll_rd_gns_upcall, - ll_wr_gns_upcall, 0 }, - - { "gns_timeout", ll_rd_gns_timeout, - ll_wr_gns_timeout, 0 }, - - { "gns_tick", ll_rd_gns_tick, - ll_wr_gns_tick, 0 }, - - { "gns_enabled", ll_rd_gns_enabled, - ll_wr_gns_enabled, 0 }, - - { "gns_object_name", ll_rd_gns_object_name, - ll_wr_gns_object_name, 0 }, - + { "max_read_ahead_mb", ll_rd_max_readahead_mb, + ll_wr_max_readahead_mb, 0 }, + { "max_read_ahead_whole_mb", ll_rd_max_read_ahead_whole_mb, + ll_wr_max_read_ahead_whole_mb, 0 }, + { "max_cached_mb", ll_rd_max_cached_mb, ll_wr_max_cached_mb, 0 }, + { "checksum_pages", ll_rd_checksum, ll_wr_checksum, 0 }, + { "max_rw_chunk", ll_rd_max_rw_chunk, ll_wr_max_rw_chunk, 0 }, + { "stats_track_pid", ll_rd_track_pid, ll_wr_track_pid, 0 }, + { "stats_track_ppid", ll_rd_track_ppid, ll_wr_track_ppid, 0 }, + { "stats_track_gid", ll_rd_track_gid, ll_wr_track_gid, 0 }, { 0 } }; @@ -509,33 +514,51 @@ struct llite_file_opcode { { LPROC_LL_FSYNC, LPROCFS_TYPE_REGS, "fsync" }, /* inode operation */ { LPROC_LL_SETATTR, LPROCFS_TYPE_REGS, "setattr" }, - { LPROC_LL_TRUNC, LPROCFS_TYPE_REGS, "punch" }, -#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0)) + { LPROC_LL_TRUNC, LPROCFS_TYPE_REGS, "truncate" }, + { LPROC_LL_FLOCK, LPROCFS_TYPE_REGS, "flock" }, { LPROC_LL_GETATTR, LPROCFS_TYPE_REGS, "getattr" }, -#else - { LPROC_LL_REVALIDATE, LPROCFS_TYPE_REGS, "getattr" }, -#endif /* special inode operation */ { LPROC_LL_STAFS, LPROCFS_TYPE_REGS, "statfs" }, { LPROC_LL_ALLOC_INODE, LPROCFS_TYPE_REGS, "alloc_inode" }, + { LPROC_LL_SETXATTR, LPROCFS_TYPE_REGS, "setxattr" }, + { LPROC_LL_GETXATTR, LPROCFS_TYPE_REGS, "getxattr" }, + { LPROC_LL_LISTXATTR, LPROCFS_TYPE_REGS, "listxattr" }, + { LPROC_LL_REMOVEXATTR, LPROCFS_TYPE_REGS, "removexattr" }, + { LPROC_LL_INODE_PERM, LPROCFS_TYPE_REGS, "inode_permission" }, { LPROC_LL_DIRECT_READ, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES, "direct_read" }, { LPROC_LL_DIRECT_WRITE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES, "direct_write" }, - { LPROC_LL_SETXATTR, LPROCFS_TYPE_REGS, "setxattr" }, - { LPROC_LL_GETXATTR, LPROCFS_TYPE_REGS, "getxattr" }, + }; +void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) +{ + if (!sbi->ll_stats) + return; + if (sbi->ll_stats_track_type == STATS_TRACK_ALL) + lprocfs_counter_add(sbi->ll_stats, op, count); + else if (sbi->ll_stats_track_type == STATS_TRACK_PID && + sbi->ll_stats_track_id == current->pid) + lprocfs_counter_add(sbi->ll_stats, op, count); + else if (sbi->ll_stats_track_type == STATS_TRACK_PPID && + sbi->ll_stats_track_id == current->p_pptr->pid) + lprocfs_counter_add(sbi->ll_stats, op, count); + else if (sbi->ll_stats_track_type == STATS_TRACK_GID && + sbi->ll_stats_track_id == current->gid) + lprocfs_counter_add(sbi->ll_stats, op, count); +} +EXPORT_SYMBOL(ll_stats_ops_tally); + int lprocfs_register_mountpoint(struct proc_dir_entry *parent, struct super_block *sb, char *osc, char *mdc) { struct lprocfs_vars lvars[2]; + struct lustre_sb_info *lsi = s2lsi(sb); struct ll_sb_info *sbi = ll_s2sbi(sb); struct obd_device *obd; - char name[MAX_STRING_SIZE + 1]; - int err, id; - struct lprocfs_stats *svc_stats = NULL; - struct proc_dir_entry *mdc_symlink, *osc_symlink; + char name[MAX_STRING_SIZE + 1], *ptr; + int err, id, len; struct proc_dir_entry *entry; ENTRY; @@ -548,10 +571,16 @@ int lprocfs_register_mountpoint(struct proc_dir_entry *parent, LASSERT(mdc != NULL); LASSERT(osc != NULL); + /* Get fsname */ + len = strlen(lsi->lsi_lmd->lmd_profile); + ptr = strrchr(lsi->lsi_lmd->lmd_profile, '-'); + if (ptr && (strcmp(ptr, "-client") == 0)) + len -= 7; + /* Mount info */ - snprintf(name, MAX_STRING_SIZE, "fs%llu", mnt_instance); - - mnt_instance++; + snprintf(name, MAX_STRING_SIZE, "%.*s-%p", len, + lsi->lsi_lmd->lmd_profile, sb); + sbi->ll_proc_root = lprocfs_register(name, parent, NULL, NULL); if (IS_ERR(sbi->ll_proc_root)) { err = PTR_ERR(sbi->ll_proc_root); @@ -565,23 +594,35 @@ int lprocfs_register_mountpoint(struct proc_dir_entry *parent, entry->proc_fops = &llite_dump_pgcache_fops; entry->data = sbi; - entry = create_proc_entry("wait_times", 0444, sbi->ll_proc_root); + entry = create_proc_entry("read_ahead_stats", 0644, sbi->ll_proc_root); if (entry == NULL) GOTO(out, err = -ENOMEM); - entry->proc_fops = &llite_wait_times_fops; + entry->proc_fops = &ll_ra_stats_fops; + entry->data = sbi; + + entry = create_proc_entry("extents_stats", 0644, sbi->ll_proc_root); + if (entry == NULL) + GOTO(out, err = -ENOMEM); + entry->proc_fops = &ll_rw_extents_stats_fops; entry->data = sbi; - entry = create_proc_entry("read_ahead_stats", 0444, sbi->ll_proc_root); + entry = create_proc_entry("extents_stats_per_process", 0644, + sbi->ll_proc_root); + if (entry == NULL) + GOTO(out, err = -ENOMEM); + entry->proc_fops = &ll_rw_extents_stats_pp_fops; + entry->data = sbi; + + entry = create_proc_entry("offset_stats", 0644, sbi->ll_proc_root); if (entry == NULL) GOTO(out, err = -ENOMEM); - entry->proc_fops = &ll_ra_stats_fops; + entry->proc_fops = &ll_rw_offset_stats_fops; entry->data = sbi; - svc_stats = lprocfs_alloc_stats(LPROC_LL_FILE_OPCODES); - if (svc_stats == NULL) { - err = -ENOMEM; - goto out; - } + /* File operations stats */ + sbi->ll_stats = lprocfs_alloc_stats(LPROC_LL_FILE_OPCODES, 0); + if (sbi->ll_stats == NULL) + GOTO(out, err = -ENOMEM); /* do counter init */ for (id = 0; id < LPROC_LL_FILE_OPCODES; id++) { __u32 type = llite_opcode_table[id].type; @@ -596,59 +637,60 @@ int lprocfs_register_mountpoint(struct proc_dir_entry *parent, ptr = "pages"; } } - lprocfs_counter_init(svc_stats, llite_opcode_table[id].opcode, + lprocfs_counter_init(sbi->ll_stats, + llite_opcode_table[id].opcode, (type & LPROCFS_CNTR_AVGMINMAX), llite_opcode_table[id].opname, ptr); } - err = lprocfs_register_stats(sbi->ll_proc_root, "stats", svc_stats); + err = lprocfs_register_stats(sbi->ll_proc_root, "stats", sbi->ll_stats); if (err) - goto out; - else - sbi->ll_stats = svc_stats; - /* need place to keep svc_stats */ + GOTO(out, err); - /* Static configuration info */ err = lprocfs_add_vars(sbi->ll_proc_root, lprocfs_obd_vars, sb); if (err) - goto out; + GOTO(out, err); /* MDC info */ obd = class_name2obd(mdc); LASSERT(obd != NULL); - LASSERT(obd->obd_type != NULL); + LASSERT(obd->obd_magic == OBD_DEVICE_MAGIC); LASSERT(obd->obd_type->typ_name != NULL); - snprintf(name, MAX_STRING_SIZE, "../../%s/%s", - obd->obd_type->typ_name, obd->obd_name); - mdc_symlink = proc_symlink(obd->obd_type->typ_name, sbi->ll_proc_root, - name); - if (mdc_symlink == NULL) { - err = -ENOMEM; - goto out; - } + snprintf(name, MAX_STRING_SIZE, "%s/common_name", + obd->obd_type->typ_name); + lvars[0].read_fptr = lprocfs_rd_name; + err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd); + if (err) + GOTO(out, err); + + snprintf(name, MAX_STRING_SIZE, "%s/uuid", obd->obd_type->typ_name); + lvars[0].read_fptr = lprocfs_rd_uuid; + err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd); + if (err) + GOTO(out, err); /* OSC */ obd = class_name2obd(osc); LASSERT(obd != NULL); - LASSERT(obd->obd_type != NULL); + LASSERT(obd->obd_magic == OBD_DEVICE_MAGIC); LASSERT(obd->obd_type->typ_name != NULL); - snprintf(name, MAX_STRING_SIZE, "../../%s/%s", - obd->obd_type->typ_name, obd->obd_name); - osc_symlink = proc_symlink(obd->obd_type->typ_name, sbi->ll_proc_root, - name); - if (osc_symlink == NULL) - err = -ENOMEM; - + snprintf(name, MAX_STRING_SIZE, "%s/common_name", + obd->obd_type->typ_name); + lvars[0].read_fptr = lprocfs_rd_name; + err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd); + if (err) + GOTO(out, err); + snprintf(name, MAX_STRING_SIZE, "%s/uuid", obd->obd_type->typ_name); + lvars[0].read_fptr = lprocfs_rd_uuid; + err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd); out: if (err) { - if (svc_stats) - lprocfs_free_stats(svc_stats); - if (sbi->ll_proc_root) - lprocfs_remove(sbi->ll_proc_root); + lprocfs_remove(&sbi->ll_proc_root); + lprocfs_free_stats(&sbi->ll_stats); } RETURN(err); } @@ -656,35 +698,12 @@ out: void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) { if (sbi->ll_proc_root) { - struct proc_dir_entry *file_stats = - lprocfs_srch(sbi->ll_proc_root, "stats"); - - if (file_stats) { - lprocfs_free_stats(sbi->ll_stats); - lprocfs_remove(file_stats); - } + lprocfs_remove(&sbi->ll_proc_root); + lprocfs_free_stats(&sbi->ll_stats); } } #undef MAX_STRING_SIZE -static struct ll_async_page *llite_pglist_next_llap(struct ll_sb_info *sbi, - struct list_head *list) -{ - struct ll_async_page *llap; - struct list_head *pos; - - list_for_each(pos, list) { - if (pos == &sbi->ll_pglist) - return NULL; - llap = list_entry(pos, struct ll_async_page, llap_proc_item); - if (llap->llap_page == NULL) - continue; - return llap; - } - LBUG(); - return NULL; -} - #define seq_page_flag(seq, page, flag, has_flags) do { \ if (test_bit(PG_##flag, &(page)->flags)) { \ if (!has_flags) \ @@ -695,6 +714,16 @@ static struct ll_async_page *llite_pglist_next_llap(struct ll_sb_info *sbi, } \ } while(0); +static void *llite_dump_pgcache_seq_start(struct seq_file *seq, loff_t *pos) +{ + struct ll_async_page *dummy_llap = seq->private; + + if (dummy_llap->llap_magic == 2) + return NULL; + + return (void *)1; +} + static int llite_dump_pgcache_seq_show(struct seq_file *seq, void *v) { struct ll_async_page *llap, *dummy_llap = seq->private; @@ -703,40 +732,42 @@ static int llite_dump_pgcache_seq_show(struct seq_file *seq, void *v) /* 2.4 doesn't seem to have SEQ_START_TOKEN, so we implement * it in our own state */ if (dummy_llap->llap_magic == 0) { - seq_printf(seq, "generation | llap cookie origin | page "); - seq_printf(seq, "inode index count [ page flags ]\n"); + seq_printf(seq, "gener | llap cookie origin wq du wb | page " + "inode index count [ page flags ]\n"); return 0; } spin_lock(&sbi->ll_lock); - llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item); + llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_pglist_item); if (llap != NULL) { int has_flags = 0; struct page *page = llap->llap_page; - static char *origins[] = { - [LLAP_ORIGIN_UNKNOWN] = "--", - [LLAP_ORIGIN_READPAGE] = "rp", - [LLAP_ORIGIN_READAHEAD] = "ra", - [LLAP_ORIGIN_COMMIT_WRITE] = "cw", - [LLAP_ORIGIN_WRITEPAGE] = "wp", - }; LASSERTF(llap->llap_origin < LLAP__ORIGIN_MAX, "%u\n", llap->llap_origin); - seq_printf(seq, "%lu | %p %p %s | %p %p %lu %u [", - sbi->ll_pglist_gen, - llap, llap->llap_cookie, - origins[llap->llap_origin], - page, page->mapping->host, page->index, - page_count(page)); + seq_printf(seq," %5lu | %p %p %s %s %s %s | %p %lu/%u(%p) " + "%lu %u [", + sbi->ll_pglist_gen, + llap, llap->llap_cookie, + llap_origins[llap->llap_origin], + llap->llap_write_queued ? "wq" : "- ", + llap->llap_defer_uptodate ? "du" : "- ", + PageWriteback(page) ? "wb" : "-", + page, page->mapping->host->i_ino, + page->mapping->host->i_generation, + page->mapping->host, page->index, + page_count(page)); seq_page_flag(seq, page, locked, has_flags); seq_page_flag(seq, page, error, has_flags); seq_page_flag(seq, page, referenced, has_flags); seq_page_flag(seq, page, uptodate, has_flags); seq_page_flag(seq, page, dirty, has_flags); +#if (LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,12)) seq_page_flag(seq, page, highmem, has_flags); +#endif + seq_page_flag(seq, page, writeback, has_flags); if (!has_flags) seq_puts(seq, "-]\n"); else @@ -748,17 +779,7 @@ static int llite_dump_pgcache_seq_show(struct seq_file *seq, void *v) return 0; } -static void *llite_dump_pgcache_seq_start(struct seq_file *seq, loff_t *pos) -{ - struct ll_async_page *llap = seq->private; - - if (llap->llap_magic == 2) - return NULL; - - return (void *)1; -} - -static void *llite_dump_pgcache_seq_next(struct seq_file *seq, void *v, +static void *llite_dump_pgcache_seq_next(struct seq_file *seq, void *v, loff_t *pos) { struct ll_async_page *llap, *dummy_llap = seq->private; @@ -774,11 +795,11 @@ static void *llite_dump_pgcache_seq_next(struct seq_file *seq, void *v, * we advance to a position beyond it, returning null if there * isn't another llap in the list beyond that new position. */ spin_lock(&sbi->ll_lock); - llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item); - list_del_init(&dummy_llap->llap_proc_item); + llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_pglist_item); + list_del_init(&dummy_llap->llap_pglist_item); if (llap) { - list_add(&dummy_llap->llap_proc_item, &llap->llap_proc_item); - llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item); + list_add(&dummy_llap->llap_pglist_item,&llap->llap_pglist_item); + llap =llite_pglist_next_llap(sbi,&dummy_llap->llap_pglist_item); } spin_unlock(&sbi->ll_lock); @@ -790,13 +811,13 @@ static void *llite_dump_pgcache_seq_next(struct seq_file *seq, void *v, return dummy_llap; } -static void llite_dump_pgcache_seq_stop(struct seq_file *seq, void *v) +static void null_stop(struct seq_file *seq, void *v) { } struct seq_operations llite_dump_pgcache_seq_sops = { .start = llite_dump_pgcache_seq_start, - .stop = llite_dump_pgcache_seq_stop, + .stop = null_stop, .next = llite_dump_pgcache_seq_next, .show = llite_dump_pgcache_seq_show, }; @@ -815,47 +836,52 @@ struct seq_operations llite_dump_pgcache_seq_sops = { static int llite_dump_pgcache_seq_open(struct inode *inode, struct file *file) { struct proc_dir_entry *dp = PDE(inode); - struct ll_async_page *llap; + struct ll_async_page *dummy_llap; struct seq_file *seq; struct ll_sb_info *sbi = dp->data; - int rc; + int rc = -ENOMEM; + + LPROCFS_ENTRY_AND_CHECK(dp); - OBD_ALLOC_GFP(llap, sizeof(*llap), GFP_KERNEL); - if (llap == NULL) - return -ENOMEM; - llap->llap_page = NULL; - llap->llap_cookie = sbi; - llap->llap_magic = 0; + OBD_ALLOC_PTR_WAIT(dummy_llap); + if (dummy_llap == NULL) + GOTO(out, rc); + dummy_llap->llap_page = NULL; + dummy_llap->llap_cookie = sbi; + dummy_llap->llap_magic = 0; rc = seq_open(file, &llite_dump_pgcache_seq_sops); if (rc) { - OBD_FREE(llap, sizeof(*llap)); - return rc; + OBD_FREE(dummy_llap, sizeof(*dummy_llap)); + GOTO(out, rc); } seq = file->private_data; - seq->private = llap; + seq->private = dummy_llap; spin_lock(&sbi->ll_lock); - list_add(&llap->llap_proc_item, &sbi->ll_pglist); + list_add(&dummy_llap->llap_pglist_item, &sbi->ll_pglist); spin_unlock(&sbi->ll_lock); - return 0; +out: + if (rc) + LPROCFS_EXIT(); + return rc; } static int llite_dump_pgcache_seq_release(struct inode *inode, struct file *file) { struct seq_file *seq = file->private_data; - struct ll_async_page *llap = seq->private; - struct ll_sb_info *sbi = llap->llap_cookie; + struct ll_async_page *dummy_llap = seq->private; + struct ll_sb_info *sbi = dummy_llap->llap_cookie; spin_lock(&sbi->ll_lock); - if (!list_empty(&llap->llap_proc_item)) - list_del_init(&llap->llap_proc_item); + if (!list_empty(&dummy_llap->llap_pglist_item)) + list_del_init(&dummy_llap->llap_pglist_item); spin_unlock(&sbi->ll_lock); - OBD_FREE(llap, sizeof(*llap)); + OBD_FREE(dummy_llap, sizeof(*dummy_llap)); - return seq_release(inode, file); + return lprocfs_seq_release(inode, file); } struct file_operations llite_dump_pgcache_fops = { @@ -864,6 +890,7 @@ struct file_operations llite_dump_pgcache_fops = { .read = seq_read, .release = llite_dump_pgcache_seq_release, }; + static int ll_ra_stats_seq_show(struct seq_file *seq, void *v) { struct timeval now; @@ -875,12 +902,14 @@ static int ll_ra_stats_seq_show(struct seq_file *seq, void *v) [RA_STAT_MISS] = "misses", [RA_STAT_DISTANT_READPAGE] = "readpage not consecutive", [RA_STAT_MISS_IN_WINDOW] = "miss inside window", + [RA_STAT_FAILED_GRAB_PAGE] = "failed grab_cache_page", [RA_STAT_FAILED_MATCH] = "failed lock match", [RA_STAT_DISCARDED] = "read but discarded", [RA_STAT_ZERO_LEN] = "zero length file", [RA_STAT_ZERO_WINDOW] = "zero size window", [RA_STAT_EOF] = "read-ahead to EOF", [RA_STAT_MAX_IN_FLIGHT] = "hit max r-a issue", + [RA_STAT_WRONG_GRAB_PAGE] = "wrong page from grab_cache_page", }; do_gettimeofday(&now); @@ -893,7 +922,7 @@ static int ll_ra_stats_seq_show(struct seq_file *seq, void *v) ra->ra_cur_pages); for(i = 0; i < _NR_RA_STAT; i++) - seq_printf(seq, "%-25s %lu\n", ra_stat_strings[i], + seq_printf(seq, "%-25s %lu\n", ra_stat_strings[i], ra->ra_stats[i]); spin_unlock(&sbi->ll_lock); @@ -901,166 +930,351 @@ static int ll_ra_stats_seq_show(struct seq_file *seq, void *v) return 0; } -static void *ll_ra_stats_seq_start(struct seq_file *p, loff_t *pos) -{ - if (*pos == 0) - return (void *)1; - return NULL; -} -static void *ll_ra_stats_seq_next(struct seq_file *p, void *v, loff_t *pos) +static ssize_t ll_ra_stats_seq_write(struct file *file, const char *buf, + size_t len, loff_t *off) { - ++*pos; - return NULL; + struct seq_file *seq = file->private_data; + struct ll_sb_info *sbi = seq->private; + struct ll_ra_info *ra = &sbi->ll_ra_info; + + spin_lock(&sbi->ll_lock); + memset(ra->ra_stats, 0, sizeof(ra->ra_stats)); + spin_unlock(&sbi->ll_lock); + + return len; } -static void ll_ra_stats_seq_stop(struct seq_file *p, void *v) + +LPROC_SEQ_FOPS(ll_ra_stats); + +#define pct(a,b) (b ? a * 100 / b : 0) + +static void ll_display_extents_info(struct ll_rw_extents_info *io_extents, + struct seq_file *seq, int which) { + unsigned long read_tot = 0, write_tot = 0, read_cum, write_cum; + unsigned long start, end, r, w; + char *unitp = "KMGTPEZY"; + int i, units = 10; + struct per_process_info *pp_info = &io_extents->pp_extents[which]; + + read_cum = 0; + write_cum = 0; + start = 0; + + for(i = 0; i < LL_HIST_MAX; i++) { + read_tot += pp_info->pp_r_hist.oh_buckets[i]; + write_tot += pp_info->pp_w_hist.oh_buckets[i]; + } + + for(i = 0; i < LL_HIST_MAX; i++) { + r = pp_info->pp_r_hist.oh_buckets[i]; + w = pp_info->pp_w_hist.oh_buckets[i]; + read_cum += r; + write_cum += w; + end = 1 << (i + LL_HIST_START - units); + seq_printf(seq, "%4lu%c - %4lu%c%c: %14lu %4lu %4lu | " + "%14lu %4lu %4lu\n", start, *unitp, end, *unitp, + (i == LL_HIST_MAX - 1) ? '+' : ' ', + r, pct(r, read_tot), pct(read_cum, read_tot), + w, pct(w, write_tot), pct(write_cum, write_tot)); + start = end; + if (start == 1<<10) { + start = 1; + units += 10; + unitp++; + } + if (read_cum == read_tot && write_cum == write_tot) + break; + } } -struct seq_operations ll_ra_stats_seq_sops = { - .start = ll_ra_stats_seq_start, - .stop = ll_ra_stats_seq_stop, - .next = ll_ra_stats_seq_next, - .show = ll_ra_stats_seq_show, -}; -static int ll_ra_stats_seq_open(struct inode *inode, struct file *file) +static int ll_rw_extents_stats_pp_seq_show(struct seq_file *seq, void *v) { - struct proc_dir_entry *dp = PDE(inode); - struct seq_file *seq; - int rc; + struct timeval now; + struct ll_sb_info *sbi = seq->private; + struct ll_rw_extents_info *io_extents = &sbi->ll_rw_extents_info; + int k; - rc = seq_open(file, &ll_ra_stats_seq_sops); - if (rc) - return rc; - seq = file->private_data; - seq->private = dp->data; + do_gettimeofday(&now); + + if (!sbi->ll_rw_stats_on) { + seq_printf(seq, "Disabled\n" + "Write anything in this file to activate\n"); + return 0; + } + seq_printf(seq, "snapshot_time: %lu.%lu (secs.usecs)\n", + now.tv_sec, now.tv_usec); + seq_printf(seq, "%15s %19s | %20s\n", " ", "read", "write"); + seq_printf(seq, "%13s %14s %4s %4s | %14s %4s %4s\n", + "extents", "calls", "%", "cum%", + "calls", "%", "cum%"); + spin_lock(&sbi->ll_pp_extent_lock); + for(k = 0; k < LL_PROCESS_HIST_MAX; k++) { + if(io_extents->pp_extents[k].pid != 0) { + seq_printf(seq, "\nPID: %d\n", + io_extents->pp_extents[k].pid); + ll_display_extents_info(io_extents, seq, k); + } + } + spin_unlock(&sbi->ll_pp_extent_lock); return 0; } -static ssize_t ll_ra_stats_seq_write(struct file *file, const char *buf, - size_t len, loff_t *off) +static ssize_t ll_rw_extents_stats_pp_seq_write(struct file *file, + const char *buf, size_t len, + loff_t *off) { struct seq_file *seq = file->private_data; struct ll_sb_info *sbi = seq->private; - struct ll_ra_info *ra = &sbi->ll_ra_info; - - spin_lock(&sbi->ll_lock); - memset(ra->ra_stats, 0, sizeof(ra->ra_stats)); - spin_unlock(&sbi->ll_lock); + struct ll_rw_extents_info *io_extents = &sbi->ll_rw_extents_info; + int i; + sbi->ll_rw_stats_on = 1; + spin_lock(&sbi->ll_pp_extent_lock); + for(i = 0; i < LL_PROCESS_HIST_MAX; i++) { + io_extents->pp_extents[i].pid = 0; + lprocfs_oh_clear(&io_extents->pp_extents[i].pp_r_hist); + lprocfs_oh_clear(&io_extents->pp_extents[i].pp_w_hist); + } + spin_unlock(&sbi->ll_pp_extent_lock); return len; } -struct file_operations ll_ra_stats_fops = { - .owner = THIS_MODULE, - .open = ll_ra_stats_seq_open, - .read = seq_read, - .write = ll_ra_stats_seq_write, - .llseek = seq_lseek, - .release = seq_release, -}; - -#define PRINTF_STIME(stime) (unsigned long)(stime)->st_num, \ - lprocfs_stime_avg_ms(stime), lprocfs_stime_avg_us(stime) +LPROC_SEQ_FOPS(ll_rw_extents_stats_pp); -static int llite_wait_times_seq_show(struct seq_file *seq, void *v) +static int ll_rw_extents_stats_seq_show(struct seq_file *seq, void *v) { - struct ll_sb_info *sbi = seq->private; struct timeval now; + struct ll_sb_info *sbi = seq->private; + struct ll_rw_extents_info *io_extents = &sbi->ll_rw_extents_info; do_gettimeofday(&now); - spin_lock(&sbi->ll_lock); - - seq_printf(seq, "snapshot_time: %lu.%lu (secs.usecs)\n\n", + if (!sbi->ll_rw_stats_on) { + seq_printf(seq, "Disabled\n" + "Write anything in this file to activate\n"); + return 0; + } + seq_printf(seq, "snapshot_time: %lu.%lu (secs.usecs)\n", now.tv_sec, now.tv_usec); - seq_printf(seq, "lock wait times: (num, average ms)\n"); - seq_printf(seq, "\tread\t%lu\t%lu.%04lu\n", - PRINTF_STIME(&sbi->ll_read_stime)); - seq_printf(seq, "\twrite\t%lu\t%lu.%04lu\n", - PRINTF_STIME(&sbi->ll_write_stime)); - seq_printf(seq, "\tgroup\t%lu\t%lu.%04lu\n", - PRINTF_STIME(&sbi->ll_grouplock_stime)); - seq_printf(seq, "\tseek\t%lu\t%lu.%04lu\n", - PRINTF_STIME(&sbi->ll_seek_stime)); - seq_printf(seq, "\tsetattr\t%lu\t%lu.%04lu\n\n", - PRINTF_STIME(&sbi->ll_setattr_stime)); - - seq_printf(seq, "io path wait times: (num, average ms)\n"); - seq_printf(seq, "\tll_brw\t%lu\t%lu.%04lu\n", - PRINTF_STIME(&sbi->ll_brw_stime)); -#if 0 - seq_printf(seq, "\tdone\t%lu\t%lu.%04lu\n", - PRINTF_STIME(&sbi->ll_done_stime)); -#endif - + seq_printf(seq, "%15s %19s | %20s\n", " ", "read", "write"); + seq_printf(seq, "%13s %14s %4s %4s | %14s %4s %4s\n", + "extents", "calls", "%", "cum%", + "calls", "%", "cum%"); + spin_lock(&sbi->ll_lock); + ll_display_extents_info(io_extents, seq, LL_PROCESS_HIST_MAX); spin_unlock(&sbi->ll_lock); return 0; } -#undef pct -static void *llite_wait_times_seq_start(struct seq_file *p, loff_t *pos) -{ - if (*pos == 0) - return (void *)1; - return NULL; -} -static void *llite_wait_times_seq_next(struct seq_file *p, void *v, loff_t *pos) +static ssize_t ll_rw_extents_stats_seq_write(struct file *file, const char *buf, + size_t len, loff_t *off) { - ++*pos; - return NULL; + struct seq_file *seq = file->private_data; + struct ll_sb_info *sbi = seq->private; + struct ll_rw_extents_info *io_extents = &sbi->ll_rw_extents_info; + int i; + + sbi->ll_rw_stats_on = 1; + spin_lock(&sbi->ll_pp_extent_lock); + for(i = 0; i <= LL_PROCESS_HIST_MAX; i++) + { + io_extents->pp_extents[i].pid = 0; + lprocfs_oh_clear(&io_extents->pp_extents[i].pp_r_hist); + lprocfs_oh_clear(&io_extents->pp_extents[i].pp_w_hist); + } + spin_unlock(&sbi->ll_pp_extent_lock); + + return len; } -static void llite_wait_times_seq_stop(struct seq_file *p, void *v) + +LPROC_SEQ_FOPS(ll_rw_extents_stats); + +void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid, struct file + *file, size_t count, int rw) { + int i, cur = -1; + struct ll_rw_process_info *process; + struct ll_rw_process_info *offset; + int *off_count = &sbi->ll_rw_offset_entry_count; + int *process_count = &sbi->ll_offset_process_count; + struct ll_rw_extents_info *io_extents = &sbi->ll_rw_extents_info; + + if(!sbi->ll_rw_stats_on) + return; + process = sbi->ll_rw_process_info; + offset = sbi->ll_rw_offset_info; + + spin_lock(&sbi->ll_pp_extent_lock); + /* Extent statistics */ + for(i = 0; i < LL_PROCESS_HIST_MAX; i++) { + if(io_extents->pp_extents[i].pid == pid) { + cur = i; + break; + } + } + + if (cur == -1) { + /* new process */ + sbi->ll_extent_process_count = + (sbi->ll_extent_process_count + 1) % LL_PROCESS_HIST_MAX; + cur = sbi->ll_extent_process_count; + io_extents->pp_extents[cur].pid = pid; + lprocfs_oh_clear(&io_extents->pp_extents[cur].pp_r_hist); + lprocfs_oh_clear(&io_extents->pp_extents[cur].pp_w_hist); + } + + for(i = 0; (count >= (1 << LL_HIST_START << i)) && + (i < (LL_HIST_MAX - 1)); i++); + if (rw == 0) { + io_extents->pp_extents[cur].pp_r_hist.oh_buckets[i]++; + io_extents->pp_extents[LL_PROCESS_HIST_MAX].pp_r_hist.oh_buckets[i]++; + } else { + io_extents->pp_extents[cur].pp_w_hist.oh_buckets[i]++; + io_extents->pp_extents[LL_PROCESS_HIST_MAX].pp_w_hist.oh_buckets[i]++; + } + spin_unlock(&sbi->ll_pp_extent_lock); + + spin_lock(&sbi->ll_process_lock); + /* Offset statistics */ + for (i = 0; i < LL_PROCESS_HIST_MAX; i++) { + if (process[i].rw_pid == pid) { + if (process[i].rw_last_file != file) { + process[i].rw_range_start = file->f_pos; + process[i].rw_last_file_pos = + file->f_pos + count; + process[i].rw_smallest_extent = count; + process[i].rw_largest_extent = count; + process[i].rw_offset = 0; + process[i].rw_last_file = file; + spin_unlock(&sbi->ll_process_lock); + return; + } + if (process[i].rw_last_file_pos != file->f_pos) { + *off_count = + (*off_count + 1) % LL_OFFSET_HIST_MAX; + offset[*off_count].rw_op = process[i].rw_op; + offset[*off_count].rw_pid = pid; + offset[*off_count].rw_range_start = + process[i].rw_range_start; + offset[*off_count].rw_range_end = + process[i].rw_last_file_pos; + offset[*off_count].rw_smallest_extent = + process[i].rw_smallest_extent; + offset[*off_count].rw_largest_extent = + process[i].rw_largest_extent; + offset[*off_count].rw_offset = + process[i].rw_offset; + process[i].rw_op = rw; + process[i].rw_range_start = file->f_pos; + process[i].rw_smallest_extent = count; + process[i].rw_largest_extent = count; + process[i].rw_offset = file->f_pos - + process[i].rw_last_file_pos; + } + if(process[i].rw_smallest_extent > count) + process[i].rw_smallest_extent = count; + if(process[i].rw_largest_extent < count) + process[i].rw_largest_extent = count; + process[i].rw_last_file_pos = file->f_pos + count; + spin_unlock(&sbi->ll_process_lock); + return; + } + } + *process_count = (*process_count + 1) % LL_PROCESS_HIST_MAX; + process[*process_count].rw_pid = pid; + process[*process_count].rw_op = rw; + process[*process_count].rw_range_start = file->f_pos; + process[*process_count].rw_last_file_pos = file->f_pos + count; + process[*process_count].rw_smallest_extent = count; + process[*process_count].rw_largest_extent = count; + process[*process_count].rw_offset = 0; + process[*process_count].rw_last_file = file; + spin_unlock(&sbi->ll_process_lock); } -struct seq_operations llite_wait_times_seq_sops = { - .start = llite_wait_times_seq_start, - .stop = llite_wait_times_seq_stop, - .next = llite_wait_times_seq_next, - .show = llite_wait_times_seq_show, -}; -static int llite_wait_times_seq_open(struct inode *inode, struct file *file) +char lpszt[] = LPSZ; + +static int ll_rw_offset_stats_seq_show(struct seq_file *seq, void *v) { - struct proc_dir_entry *dp = PDE(inode); - struct seq_file *seq; - int rc; + struct timeval now; + struct ll_sb_info *sbi = seq->private; + struct ll_rw_process_info *offset = sbi->ll_rw_offset_info; + struct ll_rw_process_info *process = sbi->ll_rw_process_info; + char format[50]; + int i; + + do_gettimeofday(&now); + + if (!sbi->ll_rw_stats_on) { + seq_printf(seq, "Disabled\n" + "Write anything in this file to activate\n"); + return 0; + } + spin_lock(&sbi->ll_process_lock); + + seq_printf(seq, "snapshot_time: %lu.%lu (secs.usecs)\n", + now.tv_sec, now.tv_usec); + seq_printf(seq, "%3s %10s %14s %14s %17s %17s %14s\n", + "R/W", "PID", "RANGE START", "RANGE END", + "SMALLEST EXTENT", "LARGEST EXTENT", "OFFSET"); + sprintf(format, "%s%s%s%s%s\n", + "%3c %10d %14Lu %14Lu %17", lpszt+1, " %17", lpszt+1, " %14Ld"); + /* We stored the discontiguous offsets here; print them first */ + for(i = 0; i < LL_OFFSET_HIST_MAX; i++) { + if (offset[i].rw_pid != 0) + /* Is there a way to snip the '%' off of LPSZ? */ + seq_printf(seq, format, + offset[i].rw_op ? 'W' : 'R', + offset[i].rw_pid, + offset[i].rw_range_start, + offset[i].rw_range_end, + offset[i].rw_smallest_extent, + offset[i].rw_largest_extent, + offset[i].rw_offset); + } + /* Then print the current offsets for each process */ + for(i = 0; i < LL_PROCESS_HIST_MAX; i++) { + if (process[i].rw_pid != 0) + seq_printf(seq, format, + process[i].rw_op ? 'W' : 'R', + process[i].rw_pid, + process[i].rw_range_start, + process[i].rw_last_file_pos, + process[i].rw_smallest_extent, + process[i].rw_largest_extent, + process[i].rw_offset); + } + spin_unlock(&sbi->ll_process_lock); - rc = seq_open(file, &llite_wait_times_seq_sops); - if (rc) - return rc; - seq = file->private_data; - seq->private = dp->data; return 0; } -static ssize_t llite_wait_times_seq_write(struct file *file, const char *buf, +static ssize_t ll_rw_offset_stats_seq_write(struct file *file, const char *buf, size_t len, loff_t *off) { struct seq_file *seq = file->private_data; struct ll_sb_info *sbi = seq->private; + struct ll_rw_process_info *process_info = sbi->ll_rw_process_info; + struct ll_rw_process_info *offset_info = sbi->ll_rw_offset_info; - spin_lock(&sbi->ll_lock); - memset(&sbi->ll_read_stime, 0, sizeof(sbi->ll_read_stime)); - memset(&sbi->ll_write_stime, 0, sizeof(sbi->ll_write_stime)); - memset(&sbi->ll_grouplock_stime, 0, sizeof(sbi->ll_grouplock_stime)); - memset(&sbi->ll_seek_stime, 0, sizeof(sbi->ll_seek_stime)); - memset(&sbi->ll_setattr_stime, 0, sizeof(sbi->ll_setattr_stime)); - memset(&sbi->ll_brw_stime, 0, sizeof(sbi->ll_brw_stime)); -// memset(&sbi->ll_done_stime, 0, sizeof(sbi->ll_done_stime)); - spin_unlock(&sbi->ll_lock); + sbi->ll_rw_stats_on = 1; + + spin_lock(&sbi->ll_process_lock); + sbi->ll_offset_process_count = 0; + sbi->ll_rw_offset_entry_count = 0; + memset(process_info, 0, sizeof(struct ll_rw_process_info) * + LL_PROCESS_HIST_MAX); + memset(offset_info, 0, sizeof(struct ll_rw_process_info) * + LL_OFFSET_HIST_MAX); + spin_unlock(&sbi->ll_process_lock); return len; } -struct file_operations llite_wait_times_fops = { - .owner = THIS_MODULE, - .open = llite_wait_times_seq_open, - .read = seq_read, - .write = llite_wait_times_seq_write, - .llseek = seq_lseek, - .release = seq_release, -}; +LPROC_SEQ_FOPS(ll_rw_offset_stats); +LPROCFS_INIT_VARS(llite, NULL, lprocfs_obd_vars) #endif /* LPROCFS */