diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2012-03-21 21:00:03 -0400 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2012-03-21 21:00:03 -0400 |
commit | ad12ab259d9131a53aa11c7c4561d97f7cc900df (patch) | |
tree | 457e99e4c4b1facf029a6a094428eaef73f5d261 | |
parent | 30d73f375238441d6dc31de0d90bc39f5bb0bfe5 (diff) | |
parent | 220cca2a4f5867db595135e0450381032eb54902 (diff) |
Merge git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-3.0-nmw
Pull gfs2 changes from Steven Whitehouse.
* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-3.0-nmw:
GFS2: Change truncate page allocation to be GFP_NOFS
GFS2: call gfs2_write_alloc_required for each chunk
GFS2: Clean up log flush header writing
GFS2: Remove a __GFP_NOFAIL allocation
GFS2: Flush pending glock work when evicting an inode
GFS2: make sure rgrps are up to date in func gfs2_blk2rgrpd
GFS2: Eliminate sd_rindex_mutex
GFS2: Unlock rindex mutex on glock error
GFS2: Make bd_cmp() static
GFS2: Sort the ordered write list
GFS2: FITRIM ioctl support
GFS2: Move two functions from log.c to lops.c
GFS2: glock statistics gathering
-rw-r--r-- | fs/gfs2/bmap.c | 4 | ||||
-rw-r--r-- | fs/gfs2/file.c | 15 | ||||
-rw-r--r-- | fs/gfs2/glock.c | 210 | ||||
-rw-r--r-- | fs/gfs2/incore.h | 50 | ||||
-rw-r--r-- | fs/gfs2/inode.c | 4 | ||||
-rw-r--r-- | fs/gfs2/lock_dlm.c | 123 | ||||
-rw-r--r-- | fs/gfs2/log.c | 244 | ||||
-rw-r--r-- | fs/gfs2/log.h | 5 | ||||
-rw-r--r-- | fs/gfs2/lops.c | 95 | ||||
-rw-r--r-- | fs/gfs2/main.c | 18 | ||||
-rw-r--r-- | fs/gfs2/ops_fstype.c | 9 | ||||
-rw-r--r-- | fs/gfs2/quota.c | 2 | ||||
-rw-r--r-- | fs/gfs2/rgrp.c | 189 | ||||
-rw-r--r-- | fs/gfs2/rgrp.h | 10 | ||||
-rw-r--r-- | fs/gfs2/super.c | 3 | ||||
-rw-r--r-- | fs/gfs2/trace_gfs2.h | 60 | ||||
-rw-r--r-- | fs/gfs2/util.c | 1 | ||||
-rw-r--r-- | fs/gfs2/util.h | 3 | ||||
-rw-r--r-- | fs/gfs2/xattr.c | 4 | ||||
-rw-r--r-- | include/linux/gfs2_ondisk.h | 1 |
20 files changed, 805 insertions, 245 deletions
diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c index 14a704015970..197c5c47e577 100644 --- a/fs/gfs2/bmap.c +++ b/fs/gfs2/bmap.c | |||
@@ -60,7 +60,7 @@ static int gfs2_unstuffer_page(struct gfs2_inode *ip, struct buffer_head *dibh, | |||
60 | int release = 0; | 60 | int release = 0; |
61 | 61 | ||
62 | if (!page || page->index) { | 62 | if (!page || page->index) { |
63 | page = grab_cache_page(inode->i_mapping, 0); | 63 | page = find_or_create_page(inode->i_mapping, 0, GFP_NOFS); |
64 | if (!page) | 64 | if (!page) |
65 | return -ENOMEM; | 65 | return -ENOMEM; |
66 | release = 1; | 66 | release = 1; |
@@ -930,7 +930,7 @@ static int gfs2_block_truncate_page(struct address_space *mapping, loff_t from) | |||
930 | struct page *page; | 930 | struct page *page; |
931 | int err; | 931 | int err; |
932 | 932 | ||
933 | page = grab_cache_page(mapping, index); | 933 | page = find_or_create_page(mapping, index, GFP_NOFS); |
934 | if (!page) | 934 | if (!page) |
935 | return 0; | 935 | return 0; |
936 | 936 | ||
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c index c5fb3597f696..76834587a8a4 100644 --- a/fs/gfs2/file.c +++ b/fs/gfs2/file.c | |||
@@ -313,6 +313,8 @@ static long gfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) | |||
313 | return gfs2_get_flags(filp, (u32 __user *)arg); | 313 | return gfs2_get_flags(filp, (u32 __user *)arg); |
314 | case FS_IOC_SETFLAGS: | 314 | case FS_IOC_SETFLAGS: |
315 | return gfs2_set_flags(filp, (u32 __user *)arg); | 315 | return gfs2_set_flags(filp, (u32 __user *)arg); |
316 | case FITRIM: | ||
317 | return gfs2_fitrim(filp, (void __user *)arg); | ||
316 | } | 318 | } |
317 | return -ENOTTY; | 319 | return -ENOTTY; |
318 | } | 320 | } |
@@ -674,6 +676,7 @@ static int fallocate_chunk(struct inode *inode, loff_t offset, loff_t len, | |||
674 | struct gfs2_inode *ip = GFS2_I(inode); | 676 | struct gfs2_inode *ip = GFS2_I(inode); |
675 | struct buffer_head *dibh; | 677 | struct buffer_head *dibh; |
676 | int error; | 678 | int error; |
679 | loff_t size = len; | ||
677 | unsigned int nr_blks; | 680 | unsigned int nr_blks; |
678 | sector_t lblock = offset >> inode->i_blkbits; | 681 | sector_t lblock = offset >> inode->i_blkbits; |
679 | 682 | ||
@@ -707,8 +710,8 @@ static int fallocate_chunk(struct inode *inode, loff_t offset, loff_t len, | |||
707 | goto out; | 710 | goto out; |
708 | } | 711 | } |
709 | } | 712 | } |
710 | if (offset + len > inode->i_size && !(mode & FALLOC_FL_KEEP_SIZE)) | 713 | if (offset + size > inode->i_size && !(mode & FALLOC_FL_KEEP_SIZE)) |
711 | i_size_write(inode, offset + len); | 714 | i_size_write(inode, offset + size); |
712 | 715 | ||
713 | mark_inode_dirty(inode); | 716 | mark_inode_dirty(inode); |
714 | 717 | ||
@@ -777,12 +780,14 @@ static long gfs2_fallocate(struct file *file, int mode, loff_t offset, | |||
777 | if (unlikely(error)) | 780 | if (unlikely(error)) |
778 | goto out_uninit; | 781 | goto out_uninit; |
779 | 782 | ||
780 | if (!gfs2_write_alloc_required(ip, offset, len)) | ||
781 | goto out_unlock; | ||
782 | |||
783 | while (len > 0) { | 783 | while (len > 0) { |
784 | if (len < bytes) | 784 | if (len < bytes) |
785 | bytes = len; | 785 | bytes = len; |
786 | if (!gfs2_write_alloc_required(ip, offset, bytes)) { | ||
787 | len -= bytes; | ||
788 | offset += bytes; | ||
789 | continue; | ||
790 | } | ||
786 | qa = gfs2_qadata_get(ip); | 791 | qa = gfs2_qadata_get(ip); |
787 | if (!qa) { | 792 | if (!qa) { |
788 | error = -ENOMEM; | 793 | error = -ENOMEM; |
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 351a3e797789..dab2526071cc 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c | |||
@@ -29,6 +29,7 @@ | |||
29 | #include <linux/rcupdate.h> | 29 | #include <linux/rcupdate.h> |
30 | #include <linux/rculist_bl.h> | 30 | #include <linux/rculist_bl.h> |
31 | #include <linux/bit_spinlock.h> | 31 | #include <linux/bit_spinlock.h> |
32 | #include <linux/percpu.h> | ||
32 | 33 | ||
33 | #include "gfs2.h" | 34 | #include "gfs2.h" |
34 | #include "incore.h" | 35 | #include "incore.h" |
@@ -543,6 +544,11 @@ __acquires(&gl->gl_spin) | |||
543 | do_error(gl, 0); /* Fail queued try locks */ | 544 | do_error(gl, 0); /* Fail queued try locks */ |
544 | } | 545 | } |
545 | gl->gl_req = target; | 546 | gl->gl_req = target; |
547 | set_bit(GLF_BLOCKING, &gl->gl_flags); | ||
548 | if ((gl->gl_req == LM_ST_UNLOCKED) || | ||
549 | (gl->gl_state == LM_ST_EXCLUSIVE) || | ||
550 | (lck_flags & (LM_FLAG_TRY|LM_FLAG_TRY_1CB))) | ||
551 | clear_bit(GLF_BLOCKING, &gl->gl_flags); | ||
546 | spin_unlock(&gl->gl_spin); | 552 | spin_unlock(&gl->gl_spin); |
547 | if (glops->go_xmote_th) | 553 | if (glops->go_xmote_th) |
548 | glops->go_xmote_th(gl); | 554 | glops->go_xmote_th(gl); |
@@ -744,6 +750,7 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number, | |||
744 | return -ENOMEM; | 750 | return -ENOMEM; |
745 | 751 | ||
746 | atomic_inc(&sdp->sd_glock_disposal); | 752 | atomic_inc(&sdp->sd_glock_disposal); |
753 | gl->gl_sbd = sdp; | ||
747 | gl->gl_flags = 0; | 754 | gl->gl_flags = 0; |
748 | gl->gl_name = name; | 755 | gl->gl_name = name; |
749 | atomic_set(&gl->gl_ref, 1); | 756 | atomic_set(&gl->gl_ref, 1); |
@@ -752,12 +759,17 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number, | |||
752 | gl->gl_demote_state = LM_ST_EXCLUSIVE; | 759 | gl->gl_demote_state = LM_ST_EXCLUSIVE; |
753 | gl->gl_hash = hash; | 760 | gl->gl_hash = hash; |
754 | gl->gl_ops = glops; | 761 | gl->gl_ops = glops; |
755 | snprintf(gl->gl_strname, GDLM_STRNAME_BYTES, "%8x%16llx", name.ln_type, (unsigned long long)number); | 762 | gl->gl_dstamp = ktime_set(0, 0); |
763 | preempt_disable(); | ||
764 | /* We use the global stats to estimate the initial per-glock stats */ | ||
765 | gl->gl_stats = this_cpu_ptr(sdp->sd_lkstats)->lkstats[glops->go_type]; | ||
766 | preempt_enable(); | ||
767 | gl->gl_stats.stats[GFS2_LKS_DCOUNT] = 0; | ||
768 | gl->gl_stats.stats[GFS2_LKS_QCOUNT] = 0; | ||
756 | memset(&gl->gl_lksb, 0, sizeof(struct dlm_lksb)); | 769 | memset(&gl->gl_lksb, 0, sizeof(struct dlm_lksb)); |
757 | gl->gl_lksb.sb_lvbptr = gl->gl_lvb; | 770 | gl->gl_lksb.sb_lvbptr = gl->gl_lvb; |
758 | gl->gl_tchange = jiffies; | 771 | gl->gl_tchange = jiffies; |
759 | gl->gl_object = NULL; | 772 | gl->gl_object = NULL; |
760 | gl->gl_sbd = sdp; | ||
761 | gl->gl_hold_time = GL_GLOCK_DFT_HOLD; | 773 | gl->gl_hold_time = GL_GLOCK_DFT_HOLD; |
762 | INIT_DELAYED_WORK(&gl->gl_work, glock_work_func); | 774 | INIT_DELAYED_WORK(&gl->gl_work, glock_work_func); |
763 | INIT_WORK(&gl->gl_delete, delete_work_func); | 775 | INIT_WORK(&gl->gl_delete, delete_work_func); |
@@ -999,6 +1011,8 @@ fail: | |||
999 | } | 1011 | } |
1000 | set_bit(GLF_QUEUED, &gl->gl_flags); | 1012 | set_bit(GLF_QUEUED, &gl->gl_flags); |
1001 | trace_gfs2_glock_queue(gh, 1); | 1013 | trace_gfs2_glock_queue(gh, 1); |
1014 | gfs2_glstats_inc(gl, GFS2_LKS_QCOUNT); | ||
1015 | gfs2_sbstats_inc(gl, GFS2_LKS_QCOUNT); | ||
1002 | if (likely(insert_pt == NULL)) { | 1016 | if (likely(insert_pt == NULL)) { |
1003 | list_add_tail(&gh->gh_list, &gl->gl_holders); | 1017 | list_add_tail(&gh->gh_list, &gl->gl_holders); |
1004 | if (unlikely(gh->gh_flags & LM_FLAG_PRIORITY)) | 1018 | if (unlikely(gh->gh_flags & LM_FLAG_PRIORITY)) |
@@ -1658,6 +1672,8 @@ static const char *gflags2str(char *buf, const struct gfs2_glock *gl) | |||
1658 | *p++ = 'L'; | 1672 | *p++ = 'L'; |
1659 | if (gl->gl_object) | 1673 | if (gl->gl_object) |
1660 | *p++ = 'o'; | 1674 | *p++ = 'o'; |
1675 | if (test_bit(GLF_BLOCKING, gflags)) | ||
1676 | *p++ = 'b'; | ||
1661 | *p = 0; | 1677 | *p = 0; |
1662 | return buf; | 1678 | return buf; |
1663 | } | 1679 | } |
@@ -1714,8 +1730,78 @@ out: | |||
1714 | return error; | 1730 | return error; |
1715 | } | 1731 | } |
1716 | 1732 | ||
1733 | static int gfs2_glstats_seq_show(struct seq_file *seq, void *iter_ptr) | ||
1734 | { | ||
1735 | struct gfs2_glock *gl = iter_ptr; | ||
1736 | |||
1737 | seq_printf(seq, "G: n:%u/%llx rtt:%lld/%lld rttb:%lld/%lld irt:%lld/%lld dcnt: %lld qcnt: %lld\n", | ||
1738 | gl->gl_name.ln_type, | ||
1739 | (unsigned long long)gl->gl_name.ln_number, | ||
1740 | (long long)gl->gl_stats.stats[GFS2_LKS_SRTT], | ||
1741 | (long long)gl->gl_stats.stats[GFS2_LKS_SRTTVAR], | ||
1742 | (long long)gl->gl_stats.stats[GFS2_LKS_SRTTB], | ||
1743 | (long long)gl->gl_stats.stats[GFS2_LKS_SRTTVARB], | ||
1744 | (long long)gl->gl_stats.stats[GFS2_LKS_SIRT], | ||
1745 | (long long)gl->gl_stats.stats[GFS2_LKS_SIRTVAR], | ||
1746 | (long long)gl->gl_stats.stats[GFS2_LKS_DCOUNT], | ||
1747 | (long long)gl->gl_stats.stats[GFS2_LKS_QCOUNT]); | ||
1748 | return 0; | ||
1749 | } | ||
1750 | |||
1751 | static const char *gfs2_gltype[] = { | ||
1752 | "type", | ||
1753 | "reserved", | ||
1754 | "nondisk", | ||
1755 | "inode", | ||
1756 | "rgrp", | ||
1757 | "meta", | ||
1758 | "iopen", | ||
1759 | "flock", | ||
1760 | "plock", | ||
1761 | "quota", | ||
1762 | "journal", | ||
1763 | }; | ||
1764 | |||
1765 | static const char *gfs2_stype[] = { | ||
1766 | [GFS2_LKS_SRTT] = "srtt", | ||
1767 | [GFS2_LKS_SRTTVAR] = "srttvar", | ||
1768 | [GFS2_LKS_SRTTB] = "srttb", | ||
1769 | [GFS2_LKS_SRTTVARB] = "srttvarb", | ||
1770 | [GFS2_LKS_SIRT] = "sirt", | ||
1771 | [GFS2_LKS_SIRTVAR] = "sirtvar", | ||
1772 | [GFS2_LKS_DCOUNT] = "dlm", | ||
1773 | [GFS2_LKS_QCOUNT] = "queue", | ||
1774 | }; | ||
1775 | |||
1776 | #define GFS2_NR_SBSTATS (ARRAY_SIZE(gfs2_gltype) * ARRAY_SIZE(gfs2_stype)) | ||
1777 | |||
1778 | static int gfs2_sbstats_seq_show(struct seq_file *seq, void *iter_ptr) | ||
1779 | { | ||
1780 | struct gfs2_glock_iter *gi = seq->private; | ||
1781 | struct gfs2_sbd *sdp = gi->sdp; | ||
1782 | unsigned index = gi->hash >> 3; | ||
1783 | unsigned subindex = gi->hash & 0x07; | ||
1784 | s64 value; | ||
1785 | int i; | ||
1786 | |||
1787 | if (index == 0 && subindex != 0) | ||
1788 | return 0; | ||
1717 | 1789 | ||
1790 | seq_printf(seq, "%-10s %8s:", gfs2_gltype[index], | ||
1791 | (index == 0) ? "cpu": gfs2_stype[subindex]); | ||
1718 | 1792 | ||
1793 | for_each_possible_cpu(i) { | ||
1794 | const struct gfs2_pcpu_lkstats *lkstats = per_cpu_ptr(sdp->sd_lkstats, i); | ||
1795 | if (index == 0) { | ||
1796 | value = i; | ||
1797 | } else { | ||
1798 | value = lkstats->lkstats[index - 1].stats[subindex]; | ||
1799 | } | ||
1800 | seq_printf(seq, " %15lld", (long long)value); | ||
1801 | } | ||
1802 | seq_putc(seq, '\n'); | ||
1803 | return 0; | ||
1804 | } | ||
1719 | 1805 | ||
1720 | int __init gfs2_glock_init(void) | 1806 | int __init gfs2_glock_init(void) |
1721 | { | 1807 | { |
@@ -1828,6 +1914,35 @@ static int gfs2_glock_seq_show(struct seq_file *seq, void *iter_ptr) | |||
1828 | return dump_glock(seq, iter_ptr); | 1914 | return dump_glock(seq, iter_ptr); |
1829 | } | 1915 | } |
1830 | 1916 | ||
1917 | static void *gfs2_sbstats_seq_start(struct seq_file *seq, loff_t *pos) | ||
1918 | { | ||
1919 | struct gfs2_glock_iter *gi = seq->private; | ||
1920 | |||
1921 | gi->hash = *pos; | ||
1922 | if (*pos >= GFS2_NR_SBSTATS) | ||
1923 | return NULL; | ||
1924 | preempt_disable(); | ||
1925 | return SEQ_START_TOKEN; | ||
1926 | } | ||
1927 | |||
1928 | static void *gfs2_sbstats_seq_next(struct seq_file *seq, void *iter_ptr, | ||
1929 | loff_t *pos) | ||
1930 | { | ||
1931 | struct gfs2_glock_iter *gi = seq->private; | ||
1932 | (*pos)++; | ||
1933 | gi->hash++; | ||
1934 | if (gi->hash >= GFS2_NR_SBSTATS) { | ||
1935 | preempt_enable(); | ||
1936 | return NULL; | ||
1937 | } | ||
1938 | return SEQ_START_TOKEN; | ||
1939 | } | ||
1940 | |||
1941 | static void gfs2_sbstats_seq_stop(struct seq_file *seq, void *iter_ptr) | ||
1942 | { | ||
1943 | preempt_enable(); | ||
1944 | } | ||
1945 | |||
1831 | static const struct seq_operations gfs2_glock_seq_ops = { | 1946 | static const struct seq_operations gfs2_glock_seq_ops = { |
1832 | .start = gfs2_glock_seq_start, | 1947 | .start = gfs2_glock_seq_start, |
1833 | .next = gfs2_glock_seq_next, | 1948 | .next = gfs2_glock_seq_next, |
@@ -1835,7 +1950,21 @@ static const struct seq_operations gfs2_glock_seq_ops = { | |||
1835 | .show = gfs2_glock_seq_show, | 1950 | .show = gfs2_glock_seq_show, |
1836 | }; | 1951 | }; |
1837 | 1952 | ||
1838 | static int gfs2_debugfs_open(struct inode *inode, struct file *file) | 1953 | static const struct seq_operations gfs2_glstats_seq_ops = { |
1954 | .start = gfs2_glock_seq_start, | ||
1955 | .next = gfs2_glock_seq_next, | ||
1956 | .stop = gfs2_glock_seq_stop, | ||
1957 | .show = gfs2_glstats_seq_show, | ||
1958 | }; | ||
1959 | |||
1960 | static const struct seq_operations gfs2_sbstats_seq_ops = { | ||
1961 | .start = gfs2_sbstats_seq_start, | ||
1962 | .next = gfs2_sbstats_seq_next, | ||
1963 | .stop = gfs2_sbstats_seq_stop, | ||
1964 | .show = gfs2_sbstats_seq_show, | ||
1965 | }; | ||
1966 | |||
1967 | static int gfs2_glocks_open(struct inode *inode, struct file *file) | ||
1839 | { | 1968 | { |
1840 | int ret = seq_open_private(file, &gfs2_glock_seq_ops, | 1969 | int ret = seq_open_private(file, &gfs2_glock_seq_ops, |
1841 | sizeof(struct gfs2_glock_iter)); | 1970 | sizeof(struct gfs2_glock_iter)); |
@@ -1847,9 +1976,49 @@ static int gfs2_debugfs_open(struct inode *inode, struct file *file) | |||
1847 | return ret; | 1976 | return ret; |
1848 | } | 1977 | } |
1849 | 1978 | ||
1850 | static const struct file_operations gfs2_debug_fops = { | 1979 | static int gfs2_glstats_open(struct inode *inode, struct file *file) |
1980 | { | ||
1981 | int ret = seq_open_private(file, &gfs2_glstats_seq_ops, | ||
1982 | sizeof(struct gfs2_glock_iter)); | ||
1983 | if (ret == 0) { | ||
1984 | struct seq_file *seq = file->private_data; | ||
1985 | struct gfs2_glock_iter *gi = seq->private; | ||
1986 | gi->sdp = inode->i_private; | ||
1987 | } | ||
1988 | return ret; | ||
1989 | } | ||
1990 | |||
1991 | static int gfs2_sbstats_open(struct inode *inode, struct file *file) | ||
1992 | { | ||
1993 | int ret = seq_open_private(file, &gfs2_sbstats_seq_ops, | ||
1994 | sizeof(struct gfs2_glock_iter)); | ||
1995 | if (ret == 0) { | ||
1996 | struct seq_file *seq = file->private_data; | ||
1997 | struct gfs2_glock_iter *gi = seq->private; | ||
1998 | gi->sdp = inode->i_private; | ||
1999 | } | ||
2000 | return ret; | ||
2001 | } | ||
2002 | |||
2003 | static const struct file_operations gfs2_glocks_fops = { | ||
2004 | .owner = THIS_MODULE, | ||
2005 | .open = gfs2_glocks_open, | ||
2006 | .read = seq_read, | ||
2007 | .llseek = seq_lseek, | ||
2008 | .release = seq_release_private, | ||
2009 | }; | ||
2010 | |||
2011 | static const struct file_operations gfs2_glstats_fops = { | ||
1851 | .owner = THIS_MODULE, | 2012 | .owner = THIS_MODULE, |
1852 | .open = gfs2_debugfs_open, | 2013 | .open = gfs2_glstats_open, |
2014 | .read = seq_read, | ||
2015 | .llseek = seq_lseek, | ||
2016 | .release = seq_release_private, | ||
2017 | }; | ||
2018 | |||
2019 | static const struct file_operations gfs2_sbstats_fops = { | ||
2020 | .owner = THIS_MODULE, | ||
2021 | .open = gfs2_sbstats_open, | ||
1853 | .read = seq_read, | 2022 | .read = seq_read, |
1854 | .llseek = seq_lseek, | 2023 | .llseek = seq_lseek, |
1855 | .release = seq_release_private, | 2024 | .release = seq_release_private, |
@@ -1863,20 +2032,45 @@ int gfs2_create_debugfs_file(struct gfs2_sbd *sdp) | |||
1863 | sdp->debugfs_dentry_glocks = debugfs_create_file("glocks", | 2032 | sdp->debugfs_dentry_glocks = debugfs_create_file("glocks", |
1864 | S_IFREG | S_IRUGO, | 2033 | S_IFREG | S_IRUGO, |
1865 | sdp->debugfs_dir, sdp, | 2034 | sdp->debugfs_dir, sdp, |
1866 | &gfs2_debug_fops); | 2035 | &gfs2_glocks_fops); |
1867 | if (!sdp->debugfs_dentry_glocks) | 2036 | if (!sdp->debugfs_dentry_glocks) |
1868 | return -ENOMEM; | 2037 | goto fail; |
2038 | |||
2039 | sdp->debugfs_dentry_glstats = debugfs_create_file("glstats", | ||
2040 | S_IFREG | S_IRUGO, | ||
2041 | sdp->debugfs_dir, sdp, | ||
2042 | &gfs2_glstats_fops); | ||
2043 | if (!sdp->debugfs_dentry_glstats) | ||
2044 | goto fail; | ||
2045 | |||
2046 | sdp->debugfs_dentry_sbstats = debugfs_create_file("sbstats", | ||
2047 | S_IFREG | S_IRUGO, | ||
2048 | sdp->debugfs_dir, sdp, | ||
2049 | &gfs2_sbstats_fops); | ||
2050 | if (!sdp->debugfs_dentry_sbstats) | ||
2051 | goto fail; | ||
1869 | 2052 | ||
1870 | return 0; | 2053 | return 0; |
2054 | fail: | ||
2055 | gfs2_delete_debugfs_file(sdp); | ||
2056 | return -ENOMEM; | ||
1871 | } | 2057 | } |
1872 | 2058 | ||
1873 | void gfs2_delete_debugfs_file(struct gfs2_sbd *sdp) | 2059 | void gfs2_delete_debugfs_file(struct gfs2_sbd *sdp) |
1874 | { | 2060 | { |
1875 | if (sdp && sdp->debugfs_dir) { | 2061 | if (sdp->debugfs_dir) { |
1876 | if (sdp->debugfs_dentry_glocks) { | 2062 | if (sdp->debugfs_dentry_glocks) { |
1877 | debugfs_remove(sdp->debugfs_dentry_glocks); | 2063 | debugfs_remove(sdp->debugfs_dentry_glocks); |
1878 | sdp->debugfs_dentry_glocks = NULL; | 2064 | sdp->debugfs_dentry_glocks = NULL; |
1879 | } | 2065 | } |
2066 | if (sdp->debugfs_dentry_glstats) { | ||
2067 | debugfs_remove(sdp->debugfs_dentry_glstats); | ||
2068 | sdp->debugfs_dentry_glstats = NULL; | ||
2069 | } | ||
2070 | if (sdp->debugfs_dentry_sbstats) { | ||
2071 | debugfs_remove(sdp->debugfs_dentry_sbstats); | ||
2072 | sdp->debugfs_dentry_sbstats = NULL; | ||
2073 | } | ||
1880 | debugfs_remove(sdp->debugfs_dir); | 2074 | debugfs_remove(sdp->debugfs_dir); |
1881 | sdp->debugfs_dir = NULL; | 2075 | sdp->debugfs_dir = NULL; |
1882 | } | 2076 | } |
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h index 97742a7ea9cc..47d0bda5ac2b 100644 --- a/fs/gfs2/incore.h +++ b/fs/gfs2/incore.h | |||
@@ -19,6 +19,8 @@ | |||
19 | #include <linux/rculist_bl.h> | 19 | #include <linux/rculist_bl.h> |
20 | #include <linux/completion.h> | 20 | #include <linux/completion.h> |
21 | #include <linux/rbtree.h> | 21 | #include <linux/rbtree.h> |
22 | #include <linux/ktime.h> | ||
23 | #include <linux/percpu.h> | ||
22 | 24 | ||
23 | #define DIO_WAIT 0x00000010 | 25 | #define DIO_WAIT 0x00000010 |
24 | #define DIO_METADATA 0x00000020 | 26 | #define DIO_METADATA 0x00000020 |
@@ -205,6 +207,22 @@ struct gfs2_glock_operations { | |||
205 | }; | 207 | }; |
206 | 208 | ||
207 | enum { | 209 | enum { |
210 | GFS2_LKS_SRTT = 0, /* Non blocking smoothed round trip time */ | ||
211 | GFS2_LKS_SRTTVAR = 1, /* Non blocking smoothed variance */ | ||
212 | GFS2_LKS_SRTTB = 2, /* Blocking smoothed round trip time */ | ||
213 | GFS2_LKS_SRTTVARB = 3, /* Blocking smoothed variance */ | ||
214 | GFS2_LKS_SIRT = 4, /* Smoothed Inter-request time */ | ||
215 | GFS2_LKS_SIRTVAR = 5, /* Smoothed Inter-request variance */ | ||
216 | GFS2_LKS_DCOUNT = 6, /* Count of dlm requests */ | ||
217 | GFS2_LKS_QCOUNT = 7, /* Count of gfs2_holder queues */ | ||
218 | GFS2_NR_LKSTATS | ||
219 | }; | ||
220 | |||
221 | struct gfs2_lkstats { | ||
222 | s64 stats[GFS2_NR_LKSTATS]; | ||
223 | }; | ||
224 | |||
225 | enum { | ||
208 | /* States */ | 226 | /* States */ |
209 | HIF_HOLDER = 6, /* Set for gh that "holds" the glock */ | 227 | HIF_HOLDER = 6, /* Set for gh that "holds" the glock */ |
210 | HIF_FIRST = 7, | 228 | HIF_FIRST = 7, |
@@ -238,10 +256,12 @@ enum { | |||
238 | GLF_QUEUED = 12, | 256 | GLF_QUEUED = 12, |
239 | GLF_LRU = 13, | 257 | GLF_LRU = 13, |
240 | GLF_OBJECT = 14, /* Used only for tracing */ | 258 | GLF_OBJECT = 14, /* Used only for tracing */ |
259 | GLF_BLOCKING = 15, | ||
241 | }; | 260 | }; |
242 | 261 | ||
243 | struct gfs2_glock { | 262 | struct gfs2_glock { |
244 | struct hlist_bl_node gl_list; | 263 | struct hlist_bl_node gl_list; |
264 | struct gfs2_sbd *gl_sbd; | ||
245 | unsigned long gl_flags; /* GLF_... */ | 265 | unsigned long gl_flags; /* GLF_... */ |
246 | struct lm_lockname gl_name; | 266 | struct lm_lockname gl_name; |
247 | atomic_t gl_ref; | 267 | atomic_t gl_ref; |
@@ -261,16 +281,14 @@ struct gfs2_glock { | |||
261 | struct list_head gl_holders; | 281 | struct list_head gl_holders; |
262 | 282 | ||
263 | const struct gfs2_glock_operations *gl_ops; | 283 | const struct gfs2_glock_operations *gl_ops; |
264 | char gl_strname[GDLM_STRNAME_BYTES]; | 284 | ktime_t gl_dstamp; |
285 | struct gfs2_lkstats gl_stats; | ||
265 | struct dlm_lksb gl_lksb; | 286 | struct dlm_lksb gl_lksb; |
266 | char gl_lvb[32]; | 287 | char gl_lvb[32]; |
267 | unsigned long gl_tchange; | 288 | unsigned long gl_tchange; |
268 | void *gl_object; | 289 | void *gl_object; |
269 | 290 | ||
270 | struct list_head gl_lru; | 291 | struct list_head gl_lru; |
271 | |||
272 | struct gfs2_sbd *gl_sbd; | ||
273 | |||
274 | struct list_head gl_ail_list; | 292 | struct list_head gl_ail_list; |
275 | atomic_t gl_ail_count; | 293 | atomic_t gl_ail_count; |
276 | atomic_t gl_revokes; | 294 | atomic_t gl_revokes; |
@@ -560,8 +578,14 @@ struct lm_lockstruct { | |||
560 | uint32_t *ls_recover_result; /* result of last jid recovery */ | 578 | uint32_t *ls_recover_result; /* result of last jid recovery */ |
561 | }; | 579 | }; |
562 | 580 | ||
581 | struct gfs2_pcpu_lkstats { | ||
582 | /* One struct for each glock type */ | ||
583 | struct gfs2_lkstats lkstats[10]; | ||
584 | }; | ||
585 | |||
563 | struct gfs2_sbd { | 586 | struct gfs2_sbd { |
564 | struct super_block *sd_vfs; | 587 | struct super_block *sd_vfs; |
588 | struct gfs2_pcpu_lkstats __percpu *sd_lkstats; | ||
565 | struct kobject sd_kobj; | 589 | struct kobject sd_kobj; |
566 | unsigned long sd_flags; /* SDF_... */ | 590 | unsigned long sd_flags; /* SDF_... */ |
567 | struct gfs2_sb_host sd_sb; | 591 | struct gfs2_sb_host sd_sb; |
@@ -620,7 +644,6 @@ struct gfs2_sbd { | |||
620 | 644 | ||
621 | int sd_rindex_uptodate; | 645 | int sd_rindex_uptodate; |
622 | spinlock_t sd_rindex_spin; | 646 | spinlock_t sd_rindex_spin; |
623 | struct mutex sd_rindex_mutex; | ||
624 | struct rb_root sd_rindex_tree; | 647 | struct rb_root sd_rindex_tree; |
625 | unsigned int sd_rgrps; | 648 | unsigned int sd_rgrps; |
626 | unsigned int sd_max_rg_data; | 649 | unsigned int sd_max_rg_data; |
@@ -725,8 +748,23 @@ struct gfs2_sbd { | |||
725 | 748 | ||
726 | unsigned long sd_last_warning; | 749 | unsigned long sd_last_warning; |
727 | struct dentry *debugfs_dir; /* debugfs directory */ | 750 | struct dentry *debugfs_dir; /* debugfs directory */ |
728 | struct dentry *debugfs_dentry_glocks; /* for debugfs */ | 751 | struct dentry *debugfs_dentry_glocks; |
752 | struct dentry *debugfs_dentry_glstats; | ||
753 | struct dentry *debugfs_dentry_sbstats; | ||
729 | }; | 754 | }; |
730 | 755 | ||
756 | static inline void gfs2_glstats_inc(struct gfs2_glock *gl, int which) | ||
757 | { | ||
758 | gl->gl_stats.stats[which]++; | ||
759 | } | ||
760 | |||
761 | static inline void gfs2_sbstats_inc(const struct gfs2_glock *gl, int which) | ||
762 | { | ||
763 | const struct gfs2_sbd *sdp = gl->gl_sbd; | ||
764 | preempt_disable(); | ||
765 | this_cpu_ptr(sdp->sd_lkstats)->lkstats[gl->gl_name.ln_type].stats[which]++; | ||
766 | preempt_enable(); | ||
767 | } | ||
768 | |||
731 | #endif /* __INCORE_DOT_H__ */ | 769 | #endif /* __INCORE_DOT_H__ */ |
732 | 770 | ||
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c index 56987460cdae..c98a60ee6dfd 100644 --- a/fs/gfs2/inode.c +++ b/fs/gfs2/inode.c | |||
@@ -1036,7 +1036,7 @@ static int gfs2_unlink(struct inode *dir, struct dentry *dentry) | |||
1036 | gfs2_holder_init(dip->i_gl, LM_ST_EXCLUSIVE, 0, ghs); | 1036 | gfs2_holder_init(dip->i_gl, LM_ST_EXCLUSIVE, 0, ghs); |
1037 | gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, ghs + 1); | 1037 | gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, ghs + 1); |
1038 | 1038 | ||
1039 | rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr); | 1039 | rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr, 1); |
1040 | if (!rgd) | 1040 | if (!rgd) |
1041 | goto out_inodes; | 1041 | goto out_inodes; |
1042 | 1042 | ||
@@ -1255,7 +1255,7 @@ static int gfs2_rename(struct inode *odir, struct dentry *odentry, | |||
1255 | * this is the case of the target file already existing | 1255 | * this is the case of the target file already existing |
1256 | * so we unlink before doing the rename | 1256 | * so we unlink before doing the rename |
1257 | */ | 1257 | */ |
1258 | nrgd = gfs2_blk2rgrpd(sdp, nip->i_no_addr); | 1258 | nrgd = gfs2_blk2rgrpd(sdp, nip->i_no_addr, 1); |
1259 | if (nrgd) | 1259 | if (nrgd) |
1260 | gfs2_holder_init(nrgd->rd_gl, LM_ST_EXCLUSIVE, 0, ghs + num_gh++); | 1260 | gfs2_holder_init(nrgd->rd_gl, LM_ST_EXCLUSIVE, 0, ghs + num_gh++); |
1261 | } | 1261 | } |
diff --git a/fs/gfs2/lock_dlm.c b/fs/gfs2/lock_dlm.c index 8944d1e32ab5..f8411bd1b805 100644 --- a/fs/gfs2/lock_dlm.c +++ b/fs/gfs2/lock_dlm.c | |||
@@ -18,14 +18,106 @@ | |||
18 | #include "glock.h" | 18 | #include "glock.h" |
19 | #include "util.h" | 19 | #include "util.h" |
20 | #include "sys.h" | 20 | #include "sys.h" |
21 | #include "trace_gfs2.h" | ||
21 | 22 | ||
22 | extern struct workqueue_struct *gfs2_control_wq; | 23 | extern struct workqueue_struct *gfs2_control_wq; |
23 | 24 | ||
25 | /** | ||
26 | * gfs2_update_stats - Update time based stats | ||
27 | * @mv: Pointer to mean/variance structure to update | ||
28 | * @sample: New data to include | ||
29 | * | ||
30 | * @delta is the difference between the current rtt sample and the | ||
31 | * running average srtt. We add 1/8 of that to the srtt in order to | ||
32 | * update the current srtt estimate. The varience estimate is a bit | ||
33 | * more complicated. We subtract the abs value of the @delta from | ||
34 | * the current variance estimate and add 1/4 of that to the running | ||
35 | * total. | ||
36 | * | ||
37 | * Note that the index points at the array entry containing the smoothed | ||
38 | * mean value, and the variance is always in the following entry | ||
39 | * | ||
40 | * Reference: TCP/IP Illustrated, vol 2, p. 831,832 | ||
41 | * All times are in units of integer nanoseconds. Unlike the TCP/IP case, | ||
42 | * they are not scaled fixed point. | ||
43 | */ | ||
44 | |||
45 | static inline void gfs2_update_stats(struct gfs2_lkstats *s, unsigned index, | ||
46 | s64 sample) | ||
47 | { | ||
48 | s64 delta = sample - s->stats[index]; | ||
49 | s->stats[index] += (delta >> 3); | ||
50 | index++; | ||
51 | s->stats[index] += ((abs64(delta) - s->stats[index]) >> 2); | ||
52 | } | ||
53 | |||
54 | /** | ||
55 | * gfs2_update_reply_times - Update locking statistics | ||
56 | * @gl: The glock to update | ||
57 | * | ||
58 | * This assumes that gl->gl_dstamp has been set earlier. | ||
59 | * | ||
60 | * The rtt (lock round trip time) is an estimate of the time | ||
61 | * taken to perform a dlm lock request. We update it on each | ||
62 | * reply from the dlm. | ||
63 | * | ||
64 | * The blocking flag is set on the glock for all dlm requests | ||
65 | * which may potentially block due to lock requests from other nodes. | ||
66 | * DLM requests where the current lock state is exclusive, the | ||
67 | * requested state is null (or unlocked) or where the TRY or | ||
68 | * TRY_1CB flags are set are classified as non-blocking. All | ||
69 | * other DLM requests are counted as (potentially) blocking. | ||
70 | */ | ||
71 | static inline void gfs2_update_reply_times(struct gfs2_glock *gl) | ||
72 | { | ||
73 | struct gfs2_pcpu_lkstats *lks; | ||
74 | const unsigned gltype = gl->gl_name.ln_type; | ||
75 | unsigned index = test_bit(GLF_BLOCKING, &gl->gl_flags) ? | ||
76 | GFS2_LKS_SRTTB : GFS2_LKS_SRTT; | ||
77 | s64 rtt; | ||
78 | |||
79 | preempt_disable(); | ||
80 | rtt = ktime_to_ns(ktime_sub(ktime_get_real(), gl->gl_dstamp)); | ||
81 | lks = this_cpu_ptr(gl->gl_sbd->sd_lkstats); | ||
82 | gfs2_update_stats(&gl->gl_stats, index, rtt); /* Local */ | ||
83 | gfs2_update_stats(&lks->lkstats[gltype], index, rtt); /* Global */ | ||
84 | preempt_enable(); | ||
85 | |||
86 | trace_gfs2_glock_lock_time(gl, rtt); | ||
87 | } | ||
88 | |||
89 | /** | ||
90 | * gfs2_update_request_times - Update locking statistics | ||
91 | * @gl: The glock to update | ||
92 | * | ||
93 | * The irt (lock inter-request times) measures the average time | ||
94 | * between requests to the dlm. It is updated immediately before | ||
95 | * each dlm call. | ||
96 | */ | ||
97 | |||
98 | static inline void gfs2_update_request_times(struct gfs2_glock *gl) | ||
99 | { | ||
100 | struct gfs2_pcpu_lkstats *lks; | ||
101 | const unsigned gltype = gl->gl_name.ln_type; | ||
102 | ktime_t dstamp; | ||
103 | s64 irt; | ||
104 | |||
105 | preempt_disable(); | ||
106 | dstamp = gl->gl_dstamp; | ||
107 | gl->gl_dstamp = ktime_get_real(); | ||
108 | irt = ktime_to_ns(ktime_sub(gl->gl_dstamp, dstamp)); | ||
109 | lks = this_cpu_ptr(gl->gl_sbd->sd_lkstats); | ||
110 | gfs2_update_stats(&gl->gl_stats, GFS2_LKS_SIRT, irt); /* Local */ | ||
111 | gfs2_update_stats(&lks->lkstats[gltype], GFS2_LKS_SIRT, irt); /* Global */ | ||
112 | preempt_enable(); | ||
113 | } | ||
114 | |||
24 | static void gdlm_ast(void *arg) | 115 | static void gdlm_ast(void *arg) |
25 | { | 116 | { |
26 | struct gfs2_glock *gl = arg; | 117 | struct gfs2_glock *gl = arg; |
27 | unsigned ret = gl->gl_state; | 118 | unsigned ret = gl->gl_state; |
28 | 119 | ||
120 | gfs2_update_reply_times(gl); | ||
29 | BUG_ON(gl->gl_lksb.sb_flags & DLM_SBF_DEMOTED); | 121 | BUG_ON(gl->gl_lksb.sb_flags & DLM_SBF_DEMOTED); |
30 | 122 | ||
31 | if (gl->gl_lksb.sb_flags & DLM_SBF_VALNOTVALID) | 123 | if (gl->gl_lksb.sb_flags & DLM_SBF_VALNOTVALID) |
@@ -111,7 +203,7 @@ static int make_mode(const unsigned int lmstate) | |||
111 | static u32 make_flags(const u32 lkid, const unsigned int gfs_flags, | 203 | static u32 make_flags(const u32 lkid, const unsigned int gfs_flags, |
112 | const int req) | 204 | const int req) |
113 | { | 205 | { |
114 | u32 lkf = 0; | 206 | u32 lkf = DLM_LKF_VALBLK; |
115 | 207 | ||
116 | if (gfs_flags & LM_FLAG_TRY) | 208 | if (gfs_flags & LM_FLAG_TRY) |
117 | lkf |= DLM_LKF_NOQUEUE; | 209 | lkf |= DLM_LKF_NOQUEUE; |
@@ -138,26 +230,43 @@ static u32 make_flags(const u32 lkid, const unsigned int gfs_flags, | |||
138 | if (lkid != 0) | 230 | if (lkid != 0) |
139 | lkf |= DLM_LKF_CONVERT; | 231 | lkf |= DLM_LKF_CONVERT; |
140 | 232 | ||
141 | lkf |= DLM_LKF_VALBLK; | ||
142 | |||
143 | return lkf; | 233 | return lkf; |
144 | } | 234 | } |
145 | 235 | ||
236 | static void gfs2_reverse_hex(char *c, u64 value) | ||
237 | { | ||
238 | while (value) { | ||
239 | *c-- = hex_asc[value & 0x0f]; | ||
240 | value >>= 4; | ||
241 | } | ||
242 | } | ||
243 | |||
146 | static int gdlm_lock(struct gfs2_glock *gl, unsigned int req_state, | 244 | static int gdlm_lock(struct gfs2_glock *gl, unsigned int req_state, |
147 | unsigned int flags) | 245 | unsigned int flags) |
148 | { | 246 | { |
149 | struct lm_lockstruct *ls = &gl->gl_sbd->sd_lockstruct; | 247 | struct lm_lockstruct *ls = &gl->gl_sbd->sd_lockstruct; |
150 | int req; | 248 | int req; |
151 | u32 lkf; | 249 | u32 lkf; |
250 | char strname[GDLM_STRNAME_BYTES] = ""; | ||
152 | 251 | ||
153 | req = make_mode(req_state); | 252 | req = make_mode(req_state); |
154 | lkf = make_flags(gl->gl_lksb.sb_lkid, flags, req); | 253 | lkf = make_flags(gl->gl_lksb.sb_lkid, flags, req); |
155 | 254 | gfs2_glstats_inc(gl, GFS2_LKS_DCOUNT); | |
255 | gfs2_sbstats_inc(gl, GFS2_LKS_DCOUNT); | ||
256 | if (gl->gl_lksb.sb_lkid) { | ||
257 | gfs2_update_request_times(gl); | ||
258 | } else { | ||
259 | memset(strname, ' ', GDLM_STRNAME_BYTES - 1); | ||
260 | strname[GDLM_STRNAME_BYTES - 1] = '\0'; | ||
261 | gfs2_reverse_hex(strname + 7, gl->gl_name.ln_type); | ||
262 | gfs2_reverse_hex(strname + 23, gl->gl_name.ln_number); | ||
263 | gl->gl_dstamp = ktime_get_real(); | ||
264 | } | ||
156 | /* | 265 | /* |
157 | * Submit the actual lock request. | 266 | * Submit the actual lock request. |
158 | */ | 267 | */ |
159 | 268 | ||
160 | return dlm_lock(ls->ls_dlm, req, &gl->gl_lksb, lkf, gl->gl_strname, | 269 | return dlm_lock(ls->ls_dlm, req, &gl->gl_lksb, lkf, strname, |
161 | GDLM_STRNAME_BYTES - 1, 0, gdlm_ast, gl, gdlm_bast); | 270 | GDLM_STRNAME_BYTES - 1, 0, gdlm_ast, gl, gdlm_bast); |
162 | } | 271 | } |
163 | 272 | ||
@@ -172,6 +281,10 @@ static void gdlm_put_lock(struct gfs2_glock *gl) | |||
172 | return; | 281 | return; |
173 | } | 282 | } |
174 | 283 | ||
284 | clear_bit(GLF_BLOCKING, &gl->gl_flags); | ||
285 | gfs2_glstats_inc(gl, GFS2_LKS_DCOUNT); | ||
286 | gfs2_sbstats_inc(gl, GFS2_LKS_DCOUNT); | ||
287 | gfs2_update_request_times(gl); | ||
175 | error = dlm_unlock(ls->ls_dlm, gl->gl_lksb.sb_lkid, DLM_LKF_VALBLK, | 288 | error = dlm_unlock(ls->ls_dlm, gl->gl_lksb.sb_lkid, DLM_LKF_VALBLK, |
176 | NULL, gl); | 289 | NULL, gl); |
177 | if (error) { | 290 | if (error) { |
diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index 756fae9eaf8f..4752eadc7f6e 100644 --- a/fs/gfs2/log.c +++ b/fs/gfs2/log.c | |||
@@ -19,6 +19,7 @@ | |||
19 | #include <linux/freezer.h> | 19 | #include <linux/freezer.h> |
20 | #include <linux/bio.h> | 20 | #include <linux/bio.h> |
21 | #include <linux/writeback.h> | 21 | #include <linux/writeback.h> |
22 | #include <linux/list_sort.h> | ||
22 | 23 | ||
23 | #include "gfs2.h" | 24 | #include "gfs2.h" |
24 | #include "incore.h" | 25 | #include "incore.h" |
@@ -358,7 +359,7 @@ retry: | |||
358 | return 0; | 359 | return 0; |
359 | } | 360 | } |
360 | 361 | ||
361 | static u64 log_bmap(struct gfs2_sbd *sdp, unsigned int lbn) | 362 | u64 gfs2_log_bmap(struct gfs2_sbd *sdp, unsigned int lbn) |
362 | { | 363 | { |
363 | struct gfs2_journal_extent *je; | 364 | struct gfs2_journal_extent *je; |
364 | 365 | ||
@@ -467,8 +468,8 @@ static unsigned int current_tail(struct gfs2_sbd *sdp) | |||
467 | 468 | ||
468 | void gfs2_log_incr_head(struct gfs2_sbd *sdp) | 469 | void gfs2_log_incr_head(struct gfs2_sbd *sdp) |
469 | { | 470 | { |
470 | if (sdp->sd_log_flush_head == sdp->sd_log_tail) | 471 | BUG_ON((sdp->sd_log_flush_head == sdp->sd_log_tail) && |
471 | BUG_ON(sdp->sd_log_flush_head != sdp->sd_log_head); | 472 | (sdp->sd_log_flush_head != sdp->sd_log_head)); |
472 | 473 | ||
473 | if (++sdp->sd_log_flush_head == sdp->sd_jdesc->jd_blocks) { | 474 | if (++sdp->sd_log_flush_head == sdp->sd_jdesc->jd_blocks) { |
474 | sdp->sd_log_flush_head = 0; | 475 | sdp->sd_log_flush_head = 0; |
@@ -476,99 +477,6 @@ void gfs2_log_incr_head(struct gfs2_sbd *sdp) | |||
476 | } | 477 | } |
477 | } | 478 | } |
478 | 479 | ||
479 | /** | ||
480 | * gfs2_log_write_endio - End of I/O for a log buffer | ||
481 | * @bh: The buffer head | ||
482 | * @uptodate: I/O Status | ||
483 | * | ||
484 | */ | ||
485 | |||
486 | static void gfs2_log_write_endio(struct buffer_head *bh, int uptodate) | ||
487 | { | ||
488 | struct gfs2_sbd *sdp = bh->b_private; | ||
489 | bh->b_private = NULL; | ||
490 | |||
491 | end_buffer_write_sync(bh, uptodate); | ||
492 | if (atomic_dec_and_test(&sdp->sd_log_in_flight)) | ||
493 | wake_up(&sdp->sd_log_flush_wait); | ||
494 | } | ||
495 | |||
496 | /** | ||
497 | * gfs2_log_get_buf - Get and initialize a buffer to use for log control data | ||
498 | * @sdp: The GFS2 superblock | ||
499 | * | ||
500 | * Returns: the buffer_head | ||
501 | */ | ||
502 | |||
503 | struct buffer_head *gfs2_log_get_buf(struct gfs2_sbd *sdp) | ||
504 | { | ||
505 | u64 blkno = log_bmap(sdp, sdp->sd_log_flush_head); | ||
506 | struct buffer_head *bh; | ||
507 | |||
508 | bh = sb_getblk(sdp->sd_vfs, blkno); | ||
509 | lock_buffer(bh); | ||
510 | memset(bh->b_data, 0, bh->b_size); | ||
511 | set_buffer_uptodate(bh); | ||
512 | clear_buffer_dirty(bh); | ||
513 | gfs2_log_incr_head(sdp); | ||
514 | atomic_inc(&sdp->sd_log_in_flight); | ||
515 | bh->b_private = sdp; | ||
516 | bh->b_end_io = gfs2_log_write_endio; | ||
517 | |||
518 | return bh; | ||
519 | } | ||
520 | |||
521 | /** | ||
522 | * gfs2_fake_write_endio - | ||
523 | * @bh: The buffer head | ||
524 | * @uptodate: The I/O Status | ||
525 | * | ||
526 | */ | ||
527 | |||
528 | static void gfs2_fake_write_endio(struct buffer_head *bh, int uptodate) | ||
529 | { | ||
530 | struct buffer_head *real_bh = bh->b_private; | ||
531 | struct gfs2_bufdata *bd = real_bh->b_private; | ||
532 | struct gfs2_sbd *sdp = bd->bd_gl->gl_sbd; | ||
533 | |||
534 | end_buffer_write_sync(bh, uptodate); | ||
535 | free_buffer_head(bh); | ||
536 | unlock_buffer(real_bh); | ||
537 | brelse(real_bh); | ||
538 | if (atomic_dec_and_test(&sdp->sd_log_in_flight)) | ||
539 | wake_up(&sdp->sd_log_flush_wait); | ||
540 | } | ||
541 | |||
542 | /** | ||
543 | * gfs2_log_fake_buf - Build a fake buffer head to write metadata buffer to log | ||
544 | * @sdp: the filesystem | ||
545 | * @data: the data the buffer_head should point to | ||
546 | * | ||
547 | * Returns: the log buffer descriptor | ||
548 | */ | ||
549 | |||
550 | struct buffer_head *gfs2_log_fake_buf(struct gfs2_sbd *sdp, | ||
551 | struct buffer_head *real) | ||
552 | { | ||
553 | u64 blkno = log_bmap(sdp, sdp->sd_log_flush_head); | ||
554 | struct buffer_head *bh; | ||
555 | |||
556 | bh = alloc_buffer_head(GFP_NOFS | __GFP_NOFAIL); | ||
557 | atomic_set(&bh->b_count, 1); | ||
558 | bh->b_state = (1 << BH_Mapped) | (1 << BH_Uptodate) | (1 << BH_Lock); | ||
559 | set_bh_page(bh, real->b_page, bh_offset(real)); | ||
560 | bh->b_blocknr = blkno; | ||
561 | bh->b_size = sdp->sd_sb.sb_bsize; | ||
562 | bh->b_bdev = sdp->sd_vfs->s_bdev; | ||
563 | bh->b_private = real; | ||
564 | bh->b_end_io = gfs2_fake_write_endio; | ||
565 | |||
566 | gfs2_log_incr_head(sdp); | ||
567 | atomic_inc(&sdp->sd_log_in_flight); | ||
568 | |||
569 | return bh; | ||
570 | } | ||
571 | |||
572 | static void log_pull_tail(struct gfs2_sbd *sdp, unsigned int new_tail) | 480 | static void log_pull_tail(struct gfs2_sbd *sdp, unsigned int new_tail) |
573 | { | 481 | { |
574 | unsigned int dist = log_distance(sdp, new_tail, sdp->sd_log_tail); | 482 | unsigned int dist = log_distance(sdp, new_tail, sdp->sd_log_tail); |
@@ -583,66 +491,8 @@ static void log_pull_tail(struct gfs2_sbd *sdp, unsigned int new_tail) | |||
583 | sdp->sd_log_tail = new_tail; | 491 | sdp->sd_log_tail = new_tail; |
584 | } | 492 | } |
585 | 493 | ||
586 | /** | ||
587 | * log_write_header - Get and initialize a journal header buffer | ||
588 | * @sdp: The GFS2 superblock | ||
589 | * | ||
590 | * Returns: the initialized log buffer descriptor | ||
591 | */ | ||
592 | 494 | ||
593 | static void log_write_header(struct gfs2_sbd *sdp, u32 flags, int pull) | 495 | static void log_flush_wait(struct gfs2_sbd *sdp) |
594 | { | ||
595 | u64 blkno = log_bmap(sdp, sdp->sd_log_flush_head); | ||
596 | struct buffer_head *bh; | ||
597 | struct gfs2_log_header *lh; | ||
598 | unsigned int tail; | ||
599 | u32 hash; | ||
600 | |||
601 | bh = sb_getblk(sdp->sd_vfs, blkno); | ||
602 | lock_buffer(bh); | ||
603 | memset(bh->b_data, 0, bh->b_size); | ||
604 | set_buffer_uptodate(bh); | ||
605 | clear_buffer_dirty(bh); | ||
606 | |||
607 | gfs2_ail1_empty(sdp); | ||
608 | tail = current_tail(sdp); | ||
609 | |||
610 | lh = (struct gfs2_log_header *)bh->b_data; | ||
611 | memset(lh, 0, sizeof(struct gfs2_log_header)); | ||
612 | lh->lh_header.mh_magic = cpu_to_be32(GFS2_MAGIC); | ||
613 | lh->lh_header.mh_type = cpu_to_be32(GFS2_METATYPE_LH); | ||
614 | lh->lh_header.__pad0 = cpu_to_be64(0); | ||
615 | lh->lh_header.mh_format = cpu_to_be32(GFS2_FORMAT_LH); | ||
616 | lh->lh_header.mh_jid = cpu_to_be32(sdp->sd_jdesc->jd_jid); | ||
617 | lh->lh_sequence = cpu_to_be64(sdp->sd_log_sequence++); | ||
618 | lh->lh_flags = cpu_to_be32(flags); | ||
619 | lh->lh_tail = cpu_to_be32(tail); | ||
620 | lh->lh_blkno = cpu_to_be32(sdp->sd_log_flush_head); | ||
621 | hash = gfs2_disk_hash(bh->b_data, sizeof(struct gfs2_log_header)); | ||
622 | lh->lh_hash = cpu_to_be32(hash); | ||
623 | |||
624 | bh->b_end_io = end_buffer_write_sync; | ||
625 | get_bh(bh); | ||
626 | if (test_bit(SDF_NOBARRIERS, &sdp->sd_flags)) | ||
627 | submit_bh(WRITE_SYNC | REQ_META | REQ_PRIO, bh); | ||
628 | else | ||
629 | submit_bh(WRITE_FLUSH_FUA | REQ_META, bh); | ||
630 | wait_on_buffer(bh); | ||
631 | |||
632 | if (!buffer_uptodate(bh)) | ||
633 | gfs2_io_error_bh(sdp, bh); | ||
634 | brelse(bh); | ||
635 | |||
636 | if (sdp->sd_log_tail != tail) | ||
637 | log_pull_tail(sdp, tail); | ||
638 | else | ||
639 | gfs2_assert_withdraw(sdp, !pull); | ||
640 | |||
641 | sdp->sd_log_idle = (tail == sdp->sd_log_flush_head); | ||
642 | gfs2_log_incr_head(sdp); | ||
643 | } | ||
644 | |||
645 | static void log_flush_commit(struct gfs2_sbd *sdp) | ||
646 | { | 496 | { |
647 | DEFINE_WAIT(wait); | 497 | DEFINE_WAIT(wait); |
648 | 498 | ||
@@ -655,8 +505,20 @@ static void log_flush_commit(struct gfs2_sbd *sdp) | |||
655 | } while(atomic_read(&sdp->sd_log_in_flight)); | 505 | } while(atomic_read(&sdp->sd_log_in_flight)); |
656 | finish_wait(&sdp->sd_log_flush_wait, &wait); | 506 | finish_wait(&sdp->sd_log_flush_wait, &wait); |
657 | } | 507 | } |
508 | } | ||
509 | |||
510 | static int bd_cmp(void *priv, struct list_head *a, struct list_head *b) | ||
511 | { | ||
512 | struct gfs2_bufdata *bda, *bdb; | ||
658 | 513 | ||
659 | log_write_header(sdp, 0, 0); | 514 | bda = list_entry(a, struct gfs2_bufdata, bd_le.le_list); |
515 | bdb = list_entry(b, struct gfs2_bufdata, bd_le.le_list); | ||
516 | |||
517 | if (bda->bd_bh->b_blocknr < bdb->bd_bh->b_blocknr) | ||
518 | return -1; | ||
519 | if (bda->bd_bh->b_blocknr > bdb->bd_bh->b_blocknr) | ||
520 | return 1; | ||
521 | return 0; | ||
660 | } | 522 | } |
661 | 523 | ||
662 | static void gfs2_ordered_write(struct gfs2_sbd *sdp) | 524 | static void gfs2_ordered_write(struct gfs2_sbd *sdp) |
@@ -666,6 +528,7 @@ static void gfs2_ordered_write(struct gfs2_sbd *sdp) | |||
666 | LIST_HEAD(written); | 528 | LIST_HEAD(written); |
667 | 529 | ||
668 | gfs2_log_lock(sdp); | 530 | gfs2_log_lock(sdp); |
531 | list_sort(NULL, &sdp->sd_log_le_ordered, &bd_cmp); | ||
669 | while (!list_empty(&sdp->sd_log_le_ordered)) { | 532 | while (!list_empty(&sdp->sd_log_le_ordered)) { |
670 | bd = list_entry(sdp->sd_log_le_ordered.next, struct gfs2_bufdata, bd_le.le_list); | 533 | bd = list_entry(sdp->sd_log_le_ordered.next, struct gfs2_bufdata, bd_le.le_list); |
671 | list_move(&bd->bd_le.le_list, &written); | 534 | list_move(&bd->bd_le.le_list, &written); |
@@ -711,6 +574,68 @@ static void gfs2_ordered_wait(struct gfs2_sbd *sdp) | |||
711 | } | 574 | } |
712 | 575 | ||
713 | /** | 576 | /** |
577 | * log_write_header - Get and initialize a journal header buffer | ||
578 | * @sdp: The GFS2 superblock | ||
579 | * | ||
580 | * Returns: the initialized log buffer descriptor | ||
581 | */ | ||
582 | |||
583 | static void log_write_header(struct gfs2_sbd *sdp, u32 flags, int pull) | ||
584 | { | ||
585 | u64 blkno = gfs2_log_bmap(sdp, sdp->sd_log_flush_head); | ||
586 | struct buffer_head *bh; | ||
587 | struct gfs2_log_header *lh; | ||
588 | unsigned int tail; | ||
589 | u32 hash; | ||
590 | |||
591 | bh = sb_getblk(sdp->sd_vfs, blkno); | ||
592 | lock_buffer(bh); | ||
593 | memset(bh->b_data, 0, bh->b_size); | ||
594 | set_buffer_uptodate(bh); | ||
595 | clear_buffer_dirty(bh); | ||
596 | |||
597 | gfs2_ail1_empty(sdp); | ||
598 | tail = current_tail(sdp); | ||
599 | |||
600 | lh = (struct gfs2_log_header *)bh->b_data; | ||
601 | memset(lh, 0, sizeof(struct gfs2_log_header)); | ||
602 | lh->lh_header.mh_magic = cpu_to_be32(GFS2_MAGIC); | ||
603 | lh->lh_header.mh_type = cpu_to_be32(GFS2_METATYPE_LH); | ||
604 | lh->lh_header.__pad0 = cpu_to_be64(0); | ||
605 | lh->lh_header.mh_format = cpu_to_be32(GFS2_FORMAT_LH); | ||
606 | lh->lh_header.mh_jid = cpu_to_be32(sdp->sd_jdesc->jd_jid); | ||
607 | lh->lh_sequence = cpu_to_be64(sdp->sd_log_sequence++); | ||
608 | lh->lh_flags = cpu_to_be32(flags); | ||
609 | lh->lh_tail = cpu_to_be32(tail); | ||
610 | lh->lh_blkno = cpu_to_be32(sdp->sd_log_flush_head); | ||
611 | hash = gfs2_disk_hash(bh->b_data, sizeof(struct gfs2_log_header)); | ||
612 | lh->lh_hash = cpu_to_be32(hash); | ||
613 | |||
614 | bh->b_end_io = end_buffer_write_sync; | ||
615 | get_bh(bh); | ||
616 | if (test_bit(SDF_NOBARRIERS, &sdp->sd_flags)) { | ||
617 | gfs2_ordered_wait(sdp); | ||
618 | log_flush_wait(sdp); | ||
619 | submit_bh(WRITE_SYNC | REQ_META | REQ_PRIO, bh); | ||
620 | } else { | ||
621 | submit_bh(WRITE_FLUSH_FUA | REQ_META, bh); | ||
622 | } | ||
623 | wait_on_buffer(bh); | ||
624 | |||
625 | if (!buffer_uptodate(bh)) | ||
626 | gfs2_io_error_bh(sdp, bh); | ||
627 | brelse(bh); | ||
628 | |||
629 | if (sdp->sd_log_tail != tail) | ||
630 | log_pull_tail(sdp, tail); | ||
631 | else | ||
632 | gfs2_assert_withdraw(sdp, !pull); | ||
633 | |||
634 | sdp->sd_log_idle = (tail == sdp->sd_log_flush_head); | ||
635 | gfs2_log_incr_head(sdp); | ||
636 | } | ||
637 | |||
638 | /** | ||
714 | * gfs2_log_flush - flush incore transaction(s) | 639 | * gfs2_log_flush - flush incore transaction(s) |
715 | * @sdp: the filesystem | 640 | * @sdp: the filesystem |
716 | * @gl: The glock structure to flush. If NULL, flush the whole incore log | 641 | * @gl: The glock structure to flush. If NULL, flush the whole incore log |
@@ -753,11 +678,10 @@ void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl) | |||
753 | 678 | ||
754 | gfs2_ordered_write(sdp); | 679 | gfs2_ordered_write(sdp); |
755 | lops_before_commit(sdp); | 680 | lops_before_commit(sdp); |
756 | gfs2_ordered_wait(sdp); | ||
757 | 681 | ||
758 | if (sdp->sd_log_head != sdp->sd_log_flush_head) | 682 | if (sdp->sd_log_head != sdp->sd_log_flush_head) { |
759 | log_flush_commit(sdp); | 683 | log_write_header(sdp, 0, 0); |
760 | else if (sdp->sd_log_tail != current_tail(sdp) && !sdp->sd_log_idle){ | 684 | } else if (sdp->sd_log_tail != current_tail(sdp) && !sdp->sd_log_idle){ |
761 | gfs2_log_lock(sdp); | 685 | gfs2_log_lock(sdp); |
762 | atomic_dec(&sdp->sd_log_blks_free); /* Adjust for unreserved buffer */ | 686 | atomic_dec(&sdp->sd_log_blks_free); /* Adjust for unreserved buffer */ |
763 | trace_gfs2_log_blocks(sdp, -1); | 687 | trace_gfs2_log_blocks(sdp, -1); |
diff --git a/fs/gfs2/log.h b/fs/gfs2/log.h index ab0621698b73..ff07454b582c 100644 --- a/fs/gfs2/log.h +++ b/fs/gfs2/log.h | |||
@@ -53,10 +53,7 @@ extern unsigned int gfs2_struct2blk(struct gfs2_sbd *sdp, unsigned int nstruct, | |||
53 | 53 | ||
54 | extern int gfs2_log_reserve(struct gfs2_sbd *sdp, unsigned int blks); | 54 | extern int gfs2_log_reserve(struct gfs2_sbd *sdp, unsigned int blks); |
55 | extern void gfs2_log_incr_head(struct gfs2_sbd *sdp); | 55 | extern void gfs2_log_incr_head(struct gfs2_sbd *sdp); |
56 | 56 | extern u64 gfs2_log_bmap(struct gfs2_sbd *sdp, unsigned int lbn); | |
57 | extern struct buffer_head *gfs2_log_get_buf(struct gfs2_sbd *sdp); | ||
58 | extern struct buffer_head *gfs2_log_fake_buf(struct gfs2_sbd *sdp, | ||
59 | struct buffer_head *real); | ||
60 | extern void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl); | 57 | extern void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl); |
61 | extern void gfs2_log_commit(struct gfs2_sbd *sdp, struct gfs2_trans *trans); | 58 | extern void gfs2_log_commit(struct gfs2_sbd *sdp, struct gfs2_trans *trans); |
62 | extern void gfs2_remove_from_ail(struct gfs2_bufdata *bd); | 59 | extern void gfs2_remove_from_ail(struct gfs2_bufdata *bd); |
diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c index df7c6e8d0764..6b1efb594d90 100644 --- a/fs/gfs2/lops.c +++ b/fs/gfs2/lops.c | |||
@@ -12,6 +12,7 @@ | |||
12 | #include <linux/spinlock.h> | 12 | #include <linux/spinlock.h> |
13 | #include <linux/completion.h> | 13 | #include <linux/completion.h> |
14 | #include <linux/buffer_head.h> | 14 | #include <linux/buffer_head.h> |
15 | #include <linux/mempool.h> | ||
15 | #include <linux/gfs2_ondisk.h> | 16 | #include <linux/gfs2_ondisk.h> |
16 | #include <linux/bio.h> | 17 | #include <linux/bio.h> |
17 | #include <linux/fs.h> | 18 | #include <linux/fs.h> |
@@ -76,7 +77,7 @@ static void maybe_release_space(struct gfs2_bufdata *bd) | |||
76 | if (bi->bi_clone == 0) | 77 | if (bi->bi_clone == 0) |
77 | return; | 78 | return; |
78 | if (sdp->sd_args.ar_discard) | 79 | if (sdp->sd_args.ar_discard) |
79 | gfs2_rgrp_send_discards(sdp, rgd->rd_data0, bd->bd_bh, bi); | 80 | gfs2_rgrp_send_discards(sdp, rgd->rd_data0, bd->bd_bh, bi, 1, NULL); |
80 | memcpy(bi->bi_clone + bi->bi_offset, | 81 | memcpy(bi->bi_clone + bi->bi_offset, |
81 | bd->bd_bh->b_data + bi->bi_offset, bi->bi_len); | 82 | bd->bd_bh->b_data + bi->bi_offset, bi->bi_len); |
82 | clear_bit(GBF_FULL, &bi->bi_flags); | 83 | clear_bit(GBF_FULL, &bi->bi_flags); |
@@ -143,6 +144,98 @@ static inline __be64 *bh_ptr_end(struct buffer_head *bh) | |||
143 | return (__force __be64 *)(bh->b_data + bh->b_size); | 144 | return (__force __be64 *)(bh->b_data + bh->b_size); |
144 | } | 145 | } |
145 | 146 | ||
147 | /** | ||
148 | * gfs2_log_write_endio - End of I/O for a log buffer | ||
149 | * @bh: The buffer head | ||
150 | * @uptodate: I/O Status | ||
151 | * | ||
152 | */ | ||
153 | |||
154 | static void gfs2_log_write_endio(struct buffer_head *bh, int uptodate) | ||
155 | { | ||
156 | struct gfs2_sbd *sdp = bh->b_private; | ||
157 | bh->b_private = NULL; | ||
158 | |||
159 | end_buffer_write_sync(bh, uptodate); | ||
160 | if (atomic_dec_and_test(&sdp->sd_log_in_flight)) | ||
161 | wake_up(&sdp->sd_log_flush_wait); | ||
162 | } | ||
163 | |||
164 | /** | ||
165 | * gfs2_log_get_buf - Get and initialize a buffer to use for log control data | ||
166 | * @sdp: The GFS2 superblock | ||
167 | * | ||
168 | * tReturns: the buffer_head | ||
169 | */ | ||
170 | |||
171 | static struct buffer_head *gfs2_log_get_buf(struct gfs2_sbd *sdp) | ||
172 | { | ||
173 | u64 blkno = gfs2_log_bmap(sdp, sdp->sd_log_flush_head); | ||
174 | struct buffer_head *bh; | ||
175 | |||
176 | bh = sb_getblk(sdp->sd_vfs, blkno); | ||
177 | lock_buffer(bh); | ||
178 | memset(bh->b_data, 0, bh->b_size); | ||
179 | set_buffer_uptodate(bh); | ||
180 | clear_buffer_dirty(bh); | ||
181 | gfs2_log_incr_head(sdp); | ||
182 | atomic_inc(&sdp->sd_log_in_flight); | ||
183 | bh->b_private = sdp; | ||
184 | bh->b_end_io = gfs2_log_write_endio; | ||
185 | |||
186 | return bh; | ||
187 | } | ||
188 | |||
189 | /** | ||
190 | * gfs2_fake_write_endio - | ||
191 | * @bh: The buffer head | ||
192 | * @uptodate: The I/O Status | ||
193 | * | ||
194 | */ | ||
195 | |||
196 | static void gfs2_fake_write_endio(struct buffer_head *bh, int uptodate) | ||
197 | { | ||
198 | struct buffer_head *real_bh = bh->b_private; | ||
199 | struct gfs2_bufdata *bd = real_bh->b_private; | ||
200 | struct gfs2_sbd *sdp = bd->bd_gl->gl_sbd; | ||
201 | |||
202 | end_buffer_write_sync(bh, uptodate); | ||
203 | mempool_free(bh, gfs2_bh_pool); | ||
204 | unlock_buffer(real_bh); | ||
205 | brelse(real_bh); | ||
206 | if (atomic_dec_and_test(&sdp->sd_log_in_flight)) | ||
207 | wake_up(&sdp->sd_log_flush_wait); | ||
208 | } | ||
209 | |||
210 | /** | ||
211 | * gfs2_log_fake_buf - Build a fake buffer head to write metadata buffer to log | ||
212 | * @sdp: the filesystem | ||
213 | * @data: the data the buffer_head should point to | ||
214 | * | ||
215 | * Returns: the log buffer descriptor | ||
216 | */ | ||
217 | |||
218 | static struct buffer_head *gfs2_log_fake_buf(struct gfs2_sbd *sdp, | ||
219 | struct buffer_head *real) | ||
220 | { | ||
221 | u64 blkno = gfs2_log_bmap(sdp, sdp->sd_log_flush_head); | ||
222 | struct buffer_head *bh; | ||
223 | |||
224 | bh = mempool_alloc(gfs2_bh_pool, GFP_NOFS); | ||
225 | atomic_set(&bh->b_count, 1); | ||
226 | bh->b_state = (1 << BH_Mapped) | (1 << BH_Uptodate) | (1 << BH_Lock); | ||
227 | set_bh_page(bh, real->b_page, bh_offset(real)); | ||
228 | bh->b_blocknr = blkno; | ||
229 | bh->b_size = sdp->sd_sb.sb_bsize; | ||
230 | bh->b_bdev = sdp->sd_vfs->s_bdev; | ||
231 | bh->b_private = real; | ||
232 | bh->b_end_io = gfs2_fake_write_endio; | ||
233 | |||
234 | gfs2_log_incr_head(sdp); | ||
235 | atomic_inc(&sdp->sd_log_in_flight); | ||
236 | |||
237 | return bh; | ||
238 | } | ||
146 | 239 | ||
147 | static struct buffer_head *gfs2_get_log_desc(struct gfs2_sbd *sdp, u32 ld_type) | 240 | static struct buffer_head *gfs2_get_log_desc(struct gfs2_sbd *sdp, u32 ld_type) |
148 | { | 241 | { |
diff --git a/fs/gfs2/main.c b/fs/gfs2/main.c index a8d9bcd0e19c..754426b1e52c 100644 --- a/fs/gfs2/main.c +++ b/fs/gfs2/main.c | |||
@@ -17,6 +17,7 @@ | |||
17 | #include <linux/rcupdate.h> | 17 | #include <linux/rcupdate.h> |
18 | #include <linux/rculist_bl.h> | 18 | #include <linux/rculist_bl.h> |
19 | #include <linux/atomic.h> | 19 | #include <linux/atomic.h> |
20 | #include <linux/mempool.h> | ||
20 | 21 | ||
21 | #include "gfs2.h" | 22 | #include "gfs2.h" |
22 | #include "incore.h" | 23 | #include "incore.h" |
@@ -69,6 +70,16 @@ static void gfs2_init_gl_aspace_once(void *foo) | |||
69 | address_space_init_once(mapping); | 70 | address_space_init_once(mapping); |
70 | } | 71 | } |
71 | 72 | ||
73 | static void *gfs2_bh_alloc(gfp_t mask, void *data) | ||
74 | { | ||
75 | return alloc_buffer_head(mask); | ||
76 | } | ||
77 | |||
78 | static void gfs2_bh_free(void *ptr, void *data) | ||
79 | { | ||
80 | return free_buffer_head(ptr); | ||
81 | } | ||
82 | |||
72 | /** | 83 | /** |
73 | * init_gfs2_fs - Register GFS2 as a filesystem | 84 | * init_gfs2_fs - Register GFS2 as a filesystem |
74 | * | 85 | * |
@@ -151,6 +162,10 @@ static int __init init_gfs2_fs(void) | |||
151 | gfs2_control_wq = alloc_workqueue("gfs2_control", | 162 | gfs2_control_wq = alloc_workqueue("gfs2_control", |
152 | WQ_NON_REENTRANT | WQ_UNBOUND | WQ_FREEZABLE, 0); | 163 | WQ_NON_REENTRANT | WQ_UNBOUND | WQ_FREEZABLE, 0); |
153 | if (!gfs2_control_wq) | 164 | if (!gfs2_control_wq) |
165 | goto fail_recovery; | ||
166 | |||
167 | gfs2_bh_pool = mempool_create(1024, gfs2_bh_alloc, gfs2_bh_free, NULL); | ||
168 | if (!gfs2_bh_pool) | ||
154 | goto fail_control; | 169 | goto fail_control; |
155 | 170 | ||
156 | gfs2_register_debugfs(); | 171 | gfs2_register_debugfs(); |
@@ -160,6 +175,8 @@ static int __init init_gfs2_fs(void) | |||
160 | return 0; | 175 | return 0; |
161 | 176 | ||
162 | fail_control: | 177 | fail_control: |
178 | destroy_workqueue(gfs2_control_wq); | ||
179 | fail_recovery: | ||
163 | destroy_workqueue(gfs_recovery_wq); | 180 | destroy_workqueue(gfs_recovery_wq); |
164 | fail_wq: | 181 | fail_wq: |
165 | unregister_filesystem(&gfs2meta_fs_type); | 182 | unregister_filesystem(&gfs2meta_fs_type); |
@@ -208,6 +225,7 @@ static void __exit exit_gfs2_fs(void) | |||
208 | 225 | ||
209 | rcu_barrier(); | 226 | rcu_barrier(); |
210 | 227 | ||
228 | mempool_destroy(gfs2_bh_pool); | ||
211 | kmem_cache_destroy(gfs2_quotad_cachep); | 229 | kmem_cache_destroy(gfs2_quotad_cachep); |
212 | kmem_cache_destroy(gfs2_rgrpd_cachep); | 230 | kmem_cache_destroy(gfs2_rgrpd_cachep); |
213 | kmem_cache_destroy(gfs2_bufdata_cachep); | 231 | kmem_cache_destroy(gfs2_bufdata_cachep); |
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 10e848c6d1b5..6f3a18f9e176 100644 --- a/fs/gfs2/ops_fstype.c +++ b/fs/gfs2/ops_fstype.c | |||
@@ -68,6 +68,12 @@ static struct gfs2_sbd *init_sbd(struct super_block *sb) | |||
68 | 68 | ||
69 | sb->s_fs_info = sdp; | 69 | sb->s_fs_info = sdp; |
70 | sdp->sd_vfs = sb; | 70 | sdp->sd_vfs = sb; |
71 | sdp->sd_lkstats = alloc_percpu(struct gfs2_pcpu_lkstats); | ||
72 | if (!sdp->sd_lkstats) { | ||
73 | kfree(sdp); | ||
74 | return NULL; | ||
75 | } | ||
76 | |||
71 | set_bit(SDF_NOJOURNALID, &sdp->sd_flags); | 77 | set_bit(SDF_NOJOURNALID, &sdp->sd_flags); |
72 | gfs2_tune_init(&sdp->sd_tune); | 78 | gfs2_tune_init(&sdp->sd_tune); |
73 | 79 | ||
@@ -77,7 +83,6 @@ static struct gfs2_sbd *init_sbd(struct super_block *sb) | |||
77 | spin_lock_init(&sdp->sd_statfs_spin); | 83 | spin_lock_init(&sdp->sd_statfs_spin); |
78 | 84 | ||
79 | spin_lock_init(&sdp->sd_rindex_spin); | 85 | spin_lock_init(&sdp->sd_rindex_spin); |
80 | mutex_init(&sdp->sd_rindex_mutex); | ||
81 | sdp->sd_rindex_tree.rb_node = NULL; | 86 | sdp->sd_rindex_tree.rb_node = NULL; |
82 | 87 | ||
83 | INIT_LIST_HEAD(&sdp->sd_jindex_list); | 88 | INIT_LIST_HEAD(&sdp->sd_jindex_list); |
@@ -1220,6 +1225,7 @@ fail_sys: | |||
1220 | gfs2_sys_fs_del(sdp); | 1225 | gfs2_sys_fs_del(sdp); |
1221 | fail: | 1226 | fail: |
1222 | gfs2_delete_debugfs_file(sdp); | 1227 | gfs2_delete_debugfs_file(sdp); |
1228 | free_percpu(sdp->sd_lkstats); | ||
1223 | kfree(sdp); | 1229 | kfree(sdp); |
1224 | sb->s_fs_info = NULL; | 1230 | sb->s_fs_info = NULL; |
1225 | return error; | 1231 | return error; |
@@ -1392,6 +1398,7 @@ static void gfs2_kill_sb(struct super_block *sb) | |||
1392 | shrink_dcache_sb(sb); | 1398 | shrink_dcache_sb(sb); |
1393 | kill_block_super(sb); | 1399 | kill_block_super(sb); |
1394 | gfs2_delete_debugfs_file(sdp); | 1400 | gfs2_delete_debugfs_file(sdp); |
1401 | free_percpu(sdp->sd_lkstats); | ||
1395 | kfree(sdp); | 1402 | kfree(sdp); |
1396 | } | 1403 | } |
1397 | 1404 | ||
diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c index c0f8904f0860..6019da3dcaed 100644 --- a/fs/gfs2/quota.c +++ b/fs/gfs2/quota.c | |||
@@ -681,7 +681,7 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc, | |||
681 | ptr = qp; | 681 | ptr = qp; |
682 | nbytes = sizeof(struct gfs2_quota); | 682 | nbytes = sizeof(struct gfs2_quota); |
683 | get_a_page: | 683 | get_a_page: |
684 | page = grab_cache_page(mapping, index); | 684 | page = find_or_create_page(mapping, index, GFP_NOFS); |
685 | if (!page) | 685 | if (!page) |
686 | return -ENOMEM; | 686 | return -ENOMEM; |
687 | 687 | ||
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c index 49ada95209d0..19bde40b4864 100644 --- a/fs/gfs2/rgrp.c +++ b/fs/gfs2/rgrp.c | |||
@@ -327,23 +327,34 @@ static inline int rgrp_contains_block(struct gfs2_rgrpd *rgd, u64 block) | |||
327 | * Returns: The resource group, or NULL if not found | 327 | * Returns: The resource group, or NULL if not found |
328 | */ | 328 | */ |
329 | 329 | ||
330 | struct gfs2_rgrpd *gfs2_blk2rgrpd(struct gfs2_sbd *sdp, u64 blk) | 330 | struct gfs2_rgrpd *gfs2_blk2rgrpd(struct gfs2_sbd *sdp, u64 blk, bool exact) |
331 | { | 331 | { |
332 | struct rb_node **newn; | 332 | struct rb_node *n, *next; |
333 | struct gfs2_rgrpd *cur; | 333 | struct gfs2_rgrpd *cur; |
334 | 334 | ||
335 | if (gfs2_rindex_update(sdp)) | ||
336 | return NULL; | ||
337 | |||
335 | spin_lock(&sdp->sd_rindex_spin); | 338 | spin_lock(&sdp->sd_rindex_spin); |
336 | newn = &sdp->sd_rindex_tree.rb_node; | 339 | n = sdp->sd_rindex_tree.rb_node; |
337 | while (*newn) { | 340 | while (n) { |
338 | cur = rb_entry(*newn, struct gfs2_rgrpd, rd_node); | 341 | cur = rb_entry(n, struct gfs2_rgrpd, rd_node); |
342 | next = NULL; | ||
339 | if (blk < cur->rd_addr) | 343 | if (blk < cur->rd_addr) |
340 | newn = &((*newn)->rb_left); | 344 | next = n->rb_left; |
341 | else if (blk >= cur->rd_data0 + cur->rd_data) | 345 | else if (blk >= cur->rd_data0 + cur->rd_data) |
342 | newn = &((*newn)->rb_right); | 346 | next = n->rb_right; |
343 | else { | 347 | if (next == NULL) { |
344 | spin_unlock(&sdp->sd_rindex_spin); | 348 | spin_unlock(&sdp->sd_rindex_spin); |
349 | if (exact) { | ||
350 | if (blk < cur->rd_addr) | ||
351 | return NULL; | ||
352 | if (blk >= cur->rd_data0 + cur->rd_data) | ||
353 | return NULL; | ||
354 | } | ||
345 | return cur; | 355 | return cur; |
346 | } | 356 | } |
357 | n = next; | ||
347 | } | 358 | } |
348 | spin_unlock(&sdp->sd_rindex_spin); | 359 | spin_unlock(&sdp->sd_rindex_spin); |
349 | 360 | ||
@@ -532,7 +543,6 @@ u64 gfs2_ri_total(struct gfs2_sbd *sdp) | |||
532 | struct file_ra_state ra_state; | 543 | struct file_ra_state ra_state; |
533 | int error, rgrps; | 544 | int error, rgrps; |
534 | 545 | ||
535 | mutex_lock(&sdp->sd_rindex_mutex); | ||
536 | file_ra_state_init(&ra_state, inode->i_mapping); | 546 | file_ra_state_init(&ra_state, inode->i_mapping); |
537 | for (rgrps = 0;; rgrps++) { | 547 | for (rgrps = 0;; rgrps++) { |
538 | loff_t pos = rgrps * sizeof(struct gfs2_rindex); | 548 | loff_t pos = rgrps * sizeof(struct gfs2_rindex); |
@@ -545,11 +555,10 @@ u64 gfs2_ri_total(struct gfs2_sbd *sdp) | |||
545 | break; | 555 | break; |
546 | total_data += be32_to_cpu(((struct gfs2_rindex *)buf)->ri_data); | 556 | total_data += be32_to_cpu(((struct gfs2_rindex *)buf)->ri_data); |
547 | } | 557 | } |
548 | mutex_unlock(&sdp->sd_rindex_mutex); | ||
549 | return total_data; | 558 | return total_data; |
550 | } | 559 | } |
551 | 560 | ||
552 | static void rgd_insert(struct gfs2_rgrpd *rgd) | 561 | static int rgd_insert(struct gfs2_rgrpd *rgd) |
553 | { | 562 | { |
554 | struct gfs2_sbd *sdp = rgd->rd_sbd; | 563 | struct gfs2_sbd *sdp = rgd->rd_sbd; |
555 | struct rb_node **newn = &sdp->sd_rindex_tree.rb_node, *parent = NULL; | 564 | struct rb_node **newn = &sdp->sd_rindex_tree.rb_node, *parent = NULL; |
@@ -565,11 +574,13 @@ static void rgd_insert(struct gfs2_rgrpd *rgd) | |||
565 | else if (rgd->rd_addr > cur->rd_addr) | 574 | else if (rgd->rd_addr > cur->rd_addr) |
566 | newn = &((*newn)->rb_right); | 575 | newn = &((*newn)->rb_right); |
567 | else | 576 | else |
568 | return; | 577 | return -EEXIST; |
569 | } | 578 | } |
570 | 579 | ||
571 | rb_link_node(&rgd->rd_node, parent, newn); | 580 | rb_link_node(&rgd->rd_node, parent, newn); |
572 | rb_insert_color(&rgd->rd_node, &sdp->sd_rindex_tree); | 581 | rb_insert_color(&rgd->rd_node, &sdp->sd_rindex_tree); |
582 | sdp->sd_rgrps++; | ||
583 | return 0; | ||
573 | } | 584 | } |
574 | 585 | ||
575 | /** | 586 | /** |
@@ -623,10 +634,12 @@ static int read_rindex_entry(struct gfs2_inode *ip, | |||
623 | if (rgd->rd_data > sdp->sd_max_rg_data) | 634 | if (rgd->rd_data > sdp->sd_max_rg_data) |
624 | sdp->sd_max_rg_data = rgd->rd_data; | 635 | sdp->sd_max_rg_data = rgd->rd_data; |
625 | spin_lock(&sdp->sd_rindex_spin); | 636 | spin_lock(&sdp->sd_rindex_spin); |
626 | rgd_insert(rgd); | 637 | error = rgd_insert(rgd); |
627 | sdp->sd_rgrps++; | ||
628 | spin_unlock(&sdp->sd_rindex_spin); | 638 | spin_unlock(&sdp->sd_rindex_spin); |
629 | return error; | 639 | if (!error) |
640 | return 0; | ||
641 | |||
642 | error = 0; /* someone else read in the rgrp; free it and ignore it */ | ||
630 | 643 | ||
631 | fail: | 644 | fail: |
632 | kfree(rgd->rd_bits); | 645 | kfree(rgd->rd_bits); |
@@ -687,7 +700,6 @@ int gfs2_rindex_update(struct gfs2_sbd *sdp) | |||
687 | 700 | ||
688 | /* Read new copy from disk if we don't have the latest */ | 701 | /* Read new copy from disk if we don't have the latest */ |
689 | if (!sdp->sd_rindex_uptodate) { | 702 | if (!sdp->sd_rindex_uptodate) { |
690 | mutex_lock(&sdp->sd_rindex_mutex); | ||
691 | if (!gfs2_glock_is_locked_by_me(gl)) { | 703 | if (!gfs2_glock_is_locked_by_me(gl)) { |
692 | error = gfs2_glock_nq_init(gl, LM_ST_SHARED, 0, &ri_gh); | 704 | error = gfs2_glock_nq_init(gl, LM_ST_SHARED, 0, &ri_gh); |
693 | if (error) | 705 | if (error) |
@@ -698,10 +710,8 @@ int gfs2_rindex_update(struct gfs2_sbd *sdp) | |||
698 | error = gfs2_ri_update(ip); | 710 | error = gfs2_ri_update(ip); |
699 | if (unlock_required) | 711 | if (unlock_required) |
700 | gfs2_glock_dq_uninit(&ri_gh); | 712 | gfs2_glock_dq_uninit(&ri_gh); |
701 | mutex_unlock(&sdp->sd_rindex_mutex); | ||
702 | } | 713 | } |
703 | 714 | ||
704 | |||
705 | return error; | 715 | return error; |
706 | } | 716 | } |
707 | 717 | ||
@@ -810,9 +820,9 @@ void gfs2_rgrp_go_unlock(struct gfs2_holder *gh) | |||
810 | 820 | ||
811 | } | 821 | } |
812 | 822 | ||
813 | void gfs2_rgrp_send_discards(struct gfs2_sbd *sdp, u64 offset, | 823 | int gfs2_rgrp_send_discards(struct gfs2_sbd *sdp, u64 offset, |
814 | struct buffer_head *bh, | 824 | struct buffer_head *bh, |
815 | const struct gfs2_bitmap *bi) | 825 | const struct gfs2_bitmap *bi, unsigned minlen, u64 *ptrimmed) |
816 | { | 826 | { |
817 | struct super_block *sb = sdp->sd_vfs; | 827 | struct super_block *sb = sdp->sd_vfs; |
818 | struct block_device *bdev = sb->s_bdev; | 828 | struct block_device *bdev = sb->s_bdev; |
@@ -823,11 +833,19 @@ void gfs2_rgrp_send_discards(struct gfs2_sbd *sdp, u64 offset, | |||
823 | sector_t nr_sects = 0; | 833 | sector_t nr_sects = 0; |
824 | int rv; | 834 | int rv; |
825 | unsigned int x; | 835 | unsigned int x; |
836 | u32 trimmed = 0; | ||
837 | u8 diff; | ||
826 | 838 | ||
827 | for (x = 0; x < bi->bi_len; x++) { | 839 | for (x = 0; x < bi->bi_len; x++) { |
828 | const u8 *orig = bh->b_data + bi->bi_offset + x; | 840 | const u8 *clone = bi->bi_clone ? bi->bi_clone : bi->bi_bh->b_data; |
829 | const u8 *clone = bi->bi_clone + bi->bi_offset + x; | 841 | clone += bi->bi_offset; |
830 | u8 diff = ~(*orig | (*orig >> 1)) & (*clone | (*clone >> 1)); | 842 | clone += x; |
843 | if (bh) { | ||
844 | const u8 *orig = bh->b_data + bi->bi_offset + x; | ||
845 | diff = ~(*orig | (*orig >> 1)) & (*clone | (*clone >> 1)); | ||
846 | } else { | ||
847 | diff = ~(*clone | (*clone >> 1)); | ||
848 | } | ||
831 | diff &= 0x55; | 849 | diff &= 0x55; |
832 | if (diff == 0) | 850 | if (diff == 0) |
833 | continue; | 851 | continue; |
@@ -838,11 +856,14 @@ void gfs2_rgrp_send_discards(struct gfs2_sbd *sdp, u64 offset, | |||
838 | if (nr_sects == 0) | 856 | if (nr_sects == 0) |
839 | goto start_new_extent; | 857 | goto start_new_extent; |
840 | if ((start + nr_sects) != blk) { | 858 | if ((start + nr_sects) != blk) { |
841 | rv = blkdev_issue_discard(bdev, start, | 859 | if (nr_sects >= minlen) { |
842 | nr_sects, GFP_NOFS, | 860 | rv = blkdev_issue_discard(bdev, |
843 | 0); | 861 | start, nr_sects, |
844 | if (rv) | 862 | GFP_NOFS, 0); |
845 | goto fail; | 863 | if (rv) |
864 | goto fail; | ||
865 | trimmed += nr_sects; | ||
866 | } | ||
846 | nr_sects = 0; | 867 | nr_sects = 0; |
847 | start_new_extent: | 868 | start_new_extent: |
848 | start = blk; | 869 | start = blk; |
@@ -853,15 +874,104 @@ start_new_extent: | |||
853 | blk += sects_per_blk; | 874 | blk += sects_per_blk; |
854 | } | 875 | } |
855 | } | 876 | } |
856 | if (nr_sects) { | 877 | if (nr_sects >= minlen) { |
857 | rv = blkdev_issue_discard(bdev, start, nr_sects, GFP_NOFS, 0); | 878 | rv = blkdev_issue_discard(bdev, start, nr_sects, GFP_NOFS, 0); |
858 | if (rv) | 879 | if (rv) |
859 | goto fail; | 880 | goto fail; |
881 | trimmed += nr_sects; | ||
860 | } | 882 | } |
861 | return; | 883 | if (ptrimmed) |
884 | *ptrimmed = trimmed; | ||
885 | return 0; | ||
886 | |||
862 | fail: | 887 | fail: |
863 | fs_warn(sdp, "error %d on discard request, turning discards off for this filesystem", rv); | 888 | if (sdp->sd_args.ar_discard) |
889 | fs_warn(sdp, "error %d on discard request, turning discards off for this filesystem", rv); | ||
864 | sdp->sd_args.ar_discard = 0; | 890 | sdp->sd_args.ar_discard = 0; |
891 | return -EIO; | ||
892 | } | ||
893 | |||
894 | /** | ||
895 | * gfs2_fitrim - Generate discard requests for unused bits of the filesystem | ||
896 | * @filp: Any file on the filesystem | ||
897 | * @argp: Pointer to the arguments (also used to pass result) | ||
898 | * | ||
899 | * Returns: 0 on success, otherwise error code | ||
900 | */ | ||
901 | |||
902 | int gfs2_fitrim(struct file *filp, void __user *argp) | ||
903 | { | ||
904 | struct inode *inode = filp->f_dentry->d_inode; | ||
905 | struct gfs2_sbd *sdp = GFS2_SB(inode); | ||
906 | struct request_queue *q = bdev_get_queue(sdp->sd_vfs->s_bdev); | ||
907 | struct buffer_head *bh; | ||
908 | struct gfs2_rgrpd *rgd; | ||
909 | struct gfs2_rgrpd *rgd_end; | ||
910 | struct gfs2_holder gh; | ||
911 | struct fstrim_range r; | ||
912 | int ret = 0; | ||
913 | u64 amt; | ||
914 | u64 trimmed = 0; | ||
915 | unsigned int x; | ||
916 | |||
917 | if (!capable(CAP_SYS_ADMIN)) | ||
918 | return -EPERM; | ||
919 | |||
920 | if (!blk_queue_discard(q)) | ||
921 | return -EOPNOTSUPP; | ||
922 | |||
923 | if (argp == NULL) { | ||
924 | r.start = 0; | ||
925 | r.len = ULLONG_MAX; | ||
926 | r.minlen = 0; | ||
927 | } else if (copy_from_user(&r, argp, sizeof(r))) | ||
928 | return -EFAULT; | ||
929 | |||
930 | rgd = gfs2_blk2rgrpd(sdp, r.start, 0); | ||
931 | rgd_end = gfs2_blk2rgrpd(sdp, r.start + r.len, 0); | ||
932 | |||
933 | while (1) { | ||
934 | |||
935 | ret = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &gh); | ||
936 | if (ret) | ||
937 | goto out; | ||
938 | |||
939 | if (!(rgd->rd_flags & GFS2_RGF_TRIMMED)) { | ||
940 | /* Trim each bitmap in the rgrp */ | ||
941 | for (x = 0; x < rgd->rd_length; x++) { | ||
942 | struct gfs2_bitmap *bi = rgd->rd_bits + x; | ||
943 | ret = gfs2_rgrp_send_discards(sdp, rgd->rd_data0, NULL, bi, r.minlen, &amt); | ||
944 | if (ret) { | ||
945 | gfs2_glock_dq_uninit(&gh); | ||
946 | goto out; | ||
947 | } | ||
948 | trimmed += amt; | ||
949 | } | ||
950 | |||
951 | /* Mark rgrp as having been trimmed */ | ||
952 | ret = gfs2_trans_begin(sdp, RES_RG_HDR, 0); | ||
953 | if (ret == 0) { | ||
954 | bh = rgd->rd_bits[0].bi_bh; | ||
955 | rgd->rd_flags |= GFS2_RGF_TRIMMED; | ||
956 | gfs2_trans_add_bh(rgd->rd_gl, bh, 1); | ||
957 | gfs2_rgrp_out(rgd, bh->b_data); | ||
958 | gfs2_trans_end(sdp); | ||
959 | } | ||
960 | } | ||
961 | gfs2_glock_dq_uninit(&gh); | ||
962 | |||
963 | if (rgd == rgd_end) | ||
964 | break; | ||
965 | |||
966 | rgd = gfs2_rgrpd_get_next(rgd); | ||
967 | } | ||
968 | |||
969 | out: | ||
970 | r.len = trimmed << 9; | ||
971 | if (argp && copy_to_user(argp, &r, sizeof(r))) | ||
972 | return -EFAULT; | ||
973 | |||
974 | return ret; | ||
865 | } | 975 | } |
866 | 976 | ||
867 | /** | 977 | /** |
@@ -1008,7 +1118,7 @@ static int get_local_rgrp(struct gfs2_inode *ip, u64 *last_unlinked) | |||
1008 | if (ip->i_rgd && rgrp_contains_block(ip->i_rgd, ip->i_goal)) | 1118 | if (ip->i_rgd && rgrp_contains_block(ip->i_rgd, ip->i_goal)) |
1009 | rgd = begin = ip->i_rgd; | 1119 | rgd = begin = ip->i_rgd; |
1010 | else | 1120 | else |
1011 | rgd = begin = gfs2_blk2rgrpd(sdp, ip->i_goal); | 1121 | rgd = begin = gfs2_blk2rgrpd(sdp, ip->i_goal, 1); |
1012 | 1122 | ||
1013 | if (rgd == NULL) | 1123 | if (rgd == NULL) |
1014 | return -EBADSLT; | 1124 | return -EBADSLT; |
@@ -1293,7 +1403,7 @@ static struct gfs2_rgrpd *rgblk_free(struct gfs2_sbd *sdp, u64 bstart, | |||
1293 | u32 length, rgrp_blk, buf_blk; | 1403 | u32 length, rgrp_blk, buf_blk; |
1294 | unsigned int buf; | 1404 | unsigned int buf; |
1295 | 1405 | ||
1296 | rgd = gfs2_blk2rgrpd(sdp, bstart); | 1406 | rgd = gfs2_blk2rgrpd(sdp, bstart, 1); |
1297 | if (!rgd) { | 1407 | if (!rgd) { |
1298 | if (gfs2_consist(sdp)) | 1408 | if (gfs2_consist(sdp)) |
1299 | fs_err(sdp, "block = %llu\n", (unsigned long long)bstart); | 1409 | fs_err(sdp, "block = %llu\n", (unsigned long long)bstart); |
@@ -1474,7 +1584,7 @@ void __gfs2_free_blocks(struct gfs2_inode *ip, u64 bstart, u32 blen, int meta) | |||
1474 | return; | 1584 | return; |
1475 | trace_gfs2_block_alloc(ip, bstart, blen, GFS2_BLKST_FREE); | 1585 | trace_gfs2_block_alloc(ip, bstart, blen, GFS2_BLKST_FREE); |
1476 | rgd->rd_free += blen; | 1586 | rgd->rd_free += blen; |
1477 | 1587 | rgd->rd_flags &= ~GFS2_RGF_TRIMMED; | |
1478 | gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1); | 1588 | gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1); |
1479 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); | 1589 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); |
1480 | 1590 | ||
@@ -1560,14 +1670,9 @@ int gfs2_check_blk_type(struct gfs2_sbd *sdp, u64 no_addr, unsigned int type) | |||
1560 | { | 1670 | { |
1561 | struct gfs2_rgrpd *rgd; | 1671 | struct gfs2_rgrpd *rgd; |
1562 | struct gfs2_holder rgd_gh; | 1672 | struct gfs2_holder rgd_gh; |
1563 | int error; | 1673 | int error = -EINVAL; |
1564 | |||
1565 | error = gfs2_rindex_update(sdp); | ||
1566 | if (error) | ||
1567 | return error; | ||
1568 | 1674 | ||
1569 | error = -EINVAL; | 1675 | rgd = gfs2_blk2rgrpd(sdp, no_addr, 1); |
1570 | rgd = gfs2_blk2rgrpd(sdp, no_addr); | ||
1571 | if (!rgd) | 1676 | if (!rgd) |
1572 | goto fail; | 1677 | goto fail; |
1573 | 1678 | ||
@@ -1610,7 +1715,7 @@ void gfs2_rlist_add(struct gfs2_inode *ip, struct gfs2_rgrp_list *rlist, | |||
1610 | if (ip->i_rgd && rgrp_contains_block(ip->i_rgd, block)) | 1715 | if (ip->i_rgd && rgrp_contains_block(ip->i_rgd, block)) |
1611 | rgd = ip->i_rgd; | 1716 | rgd = ip->i_rgd; |
1612 | else | 1717 | else |
1613 | rgd = gfs2_blk2rgrpd(sdp, block); | 1718 | rgd = gfs2_blk2rgrpd(sdp, block, 1); |
1614 | if (!rgd) { | 1719 | if (!rgd) { |
1615 | fs_err(sdp, "rlist_add: no rgrp for block %llu\n", (unsigned long long)block); | 1720 | fs_err(sdp, "rlist_add: no rgrp for block %llu\n", (unsigned long long)block); |
1616 | return; | 1721 | return; |
diff --git a/fs/gfs2/rgrp.h b/fs/gfs2/rgrp.h index ceec9106cdf4..b4b10f4de25f 100644 --- a/fs/gfs2/rgrp.h +++ b/fs/gfs2/rgrp.h | |||
@@ -11,6 +11,7 @@ | |||
11 | #define __RGRP_DOT_H__ | 11 | #define __RGRP_DOT_H__ |
12 | 12 | ||
13 | #include <linux/slab.h> | 13 | #include <linux/slab.h> |
14 | #include <linux/uaccess.h> | ||
14 | 15 | ||
15 | struct gfs2_rgrpd; | 16 | struct gfs2_rgrpd; |
16 | struct gfs2_sbd; | 17 | struct gfs2_sbd; |
@@ -18,7 +19,7 @@ struct gfs2_holder; | |||
18 | 19 | ||
19 | extern void gfs2_rgrp_verify(struct gfs2_rgrpd *rgd); | 20 | extern void gfs2_rgrp_verify(struct gfs2_rgrpd *rgd); |
20 | 21 | ||
21 | extern struct gfs2_rgrpd *gfs2_blk2rgrpd(struct gfs2_sbd *sdp, u64 blk); | 22 | extern struct gfs2_rgrpd *gfs2_blk2rgrpd(struct gfs2_sbd *sdp, u64 blk, bool exact); |
22 | extern struct gfs2_rgrpd *gfs2_rgrpd_get_first(struct gfs2_sbd *sdp); | 23 | extern struct gfs2_rgrpd *gfs2_rgrpd_get_first(struct gfs2_sbd *sdp); |
23 | extern struct gfs2_rgrpd *gfs2_rgrpd_get_next(struct gfs2_rgrpd *rgd); | 24 | extern struct gfs2_rgrpd *gfs2_rgrpd_get_next(struct gfs2_rgrpd *rgd); |
24 | 25 | ||
@@ -62,8 +63,9 @@ extern void gfs2_rlist_alloc(struct gfs2_rgrp_list *rlist, unsigned int state); | |||
62 | extern void gfs2_rlist_free(struct gfs2_rgrp_list *rlist); | 63 | extern void gfs2_rlist_free(struct gfs2_rgrp_list *rlist); |
63 | extern u64 gfs2_ri_total(struct gfs2_sbd *sdp); | 64 | extern u64 gfs2_ri_total(struct gfs2_sbd *sdp); |
64 | extern int gfs2_rgrp_dump(struct seq_file *seq, const struct gfs2_glock *gl); | 65 | extern int gfs2_rgrp_dump(struct seq_file *seq, const struct gfs2_glock *gl); |
65 | extern void gfs2_rgrp_send_discards(struct gfs2_sbd *sdp, u64 offset, | 66 | extern int gfs2_rgrp_send_discards(struct gfs2_sbd *sdp, u64 offset, |
66 | struct buffer_head *bh, | 67 | struct buffer_head *bh, |
67 | const struct gfs2_bitmap *bi); | 68 | const struct gfs2_bitmap *bi, unsigned minlen, u64 *ptrimmed); |
69 | extern int gfs2_fitrim(struct file *filp, void __user *argp); | ||
68 | 70 | ||
69 | #endif /* __RGRP_DOT_H__ */ | 71 | #endif /* __RGRP_DOT_H__ */ |
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c index 4553ce515f62..6172fa77ad59 100644 --- a/fs/gfs2/super.c +++ b/fs/gfs2/super.c | |||
@@ -1417,7 +1417,7 @@ static int gfs2_dinode_dealloc(struct gfs2_inode *ip) | |||
1417 | if (error) | 1417 | if (error) |
1418 | goto out; | 1418 | goto out; |
1419 | 1419 | ||
1420 | rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr); | 1420 | rgd = gfs2_blk2rgrpd(sdp, ip->i_no_addr, 1); |
1421 | if (!rgd) { | 1421 | if (!rgd) { |
1422 | gfs2_consist_inode(ip); | 1422 | gfs2_consist_inode(ip); |
1423 | error = -EIO; | 1423 | error = -EIO; |
@@ -1557,6 +1557,7 @@ out: | |||
1557 | end_writeback(inode); | 1557 | end_writeback(inode); |
1558 | gfs2_dir_hash_inval(ip); | 1558 | gfs2_dir_hash_inval(ip); |
1559 | ip->i_gl->gl_object = NULL; | 1559 | ip->i_gl->gl_object = NULL; |
1560 | flush_delayed_work_sync(&ip->i_gl->gl_work); | ||
1560 | gfs2_glock_add_to_lru(ip->i_gl); | 1561 | gfs2_glock_add_to_lru(ip->i_gl); |
1561 | gfs2_glock_put(ip->i_gl); | 1562 | gfs2_glock_put(ip->i_gl); |
1562 | ip->i_gl = NULL; | 1563 | ip->i_gl = NULL; |
diff --git a/fs/gfs2/trace_gfs2.h b/fs/gfs2/trace_gfs2.h index 5d07609ec57d..dfa89cd75534 100644 --- a/fs/gfs2/trace_gfs2.h +++ b/fs/gfs2/trace_gfs2.h | |||
@@ -11,6 +11,7 @@ | |||
11 | #include <linux/dlmconstants.h> | 11 | #include <linux/dlmconstants.h> |
12 | #include <linux/gfs2_ondisk.h> | 12 | #include <linux/gfs2_ondisk.h> |
13 | #include <linux/writeback.h> | 13 | #include <linux/writeback.h> |
14 | #include <linux/ktime.h> | ||
14 | #include "incore.h" | 15 | #include "incore.h" |
15 | #include "glock.h" | 16 | #include "glock.h" |
16 | 17 | ||
@@ -43,7 +44,8 @@ | |||
43 | {(1UL << GLF_FROZEN), "F" }, \ | 44 | {(1UL << GLF_FROZEN), "F" }, \ |
44 | {(1UL << GLF_QUEUED), "q" }, \ | 45 | {(1UL << GLF_QUEUED), "q" }, \ |
45 | {(1UL << GLF_LRU), "L" }, \ | 46 | {(1UL << GLF_LRU), "L" }, \ |
46 | {(1UL << GLF_OBJECT), "o" }) | 47 | {(1UL << GLF_OBJECT), "o" }, \ |
48 | {(1UL << GLF_BLOCKING), "b" }) | ||
47 | 49 | ||
48 | #ifndef NUMPTY | 50 | #ifndef NUMPTY |
49 | #define NUMPTY | 51 | #define NUMPTY |
@@ -236,6 +238,62 @@ TRACE_EVENT(gfs2_glock_queue, | |||
236 | glock_trace_name(__entry->state)) | 238 | glock_trace_name(__entry->state)) |
237 | ); | 239 | ); |
238 | 240 | ||
241 | /* DLM sends a reply to GFS2 */ | ||
242 | TRACE_EVENT(gfs2_glock_lock_time, | ||
243 | |||
244 | TP_PROTO(const struct gfs2_glock *gl, s64 tdiff), | ||
245 | |||
246 | TP_ARGS(gl, tdiff), | ||
247 | |||
248 | TP_STRUCT__entry( | ||
249 | __field( dev_t, dev ) | ||
250 | __field( u64, glnum ) | ||
251 | __field( u32, gltype ) | ||
252 | __field( int, status ) | ||
253 | __field( char, flags ) | ||
254 | __field( s64, tdiff ) | ||
255 | __field( s64, srtt ) | ||
256 | __field( s64, srttvar ) | ||
257 | __field( s64, srttb ) | ||
258 | __field( s64, srttvarb ) | ||
259 | __field( s64, sirt ) | ||
260 | __field( s64, sirtvar ) | ||
261 | __field( s64, dcount ) | ||
262 | __field( s64, qcount ) | ||
263 | ), | ||
264 | |||
265 | TP_fast_assign( | ||
266 | __entry->dev = gl->gl_sbd->sd_vfs->s_dev; | ||
267 | __entry->glnum = gl->gl_name.ln_number; | ||
268 | __entry->gltype = gl->gl_name.ln_type; | ||
269 | __entry->status = gl->gl_lksb.sb_status; | ||
270 | __entry->flags = gl->gl_lksb.sb_flags; | ||
271 | __entry->tdiff = tdiff; | ||
272 | __entry->srtt = gl->gl_stats.stats[GFS2_LKS_SRTT]; | ||
273 | __entry->srttvar = gl->gl_stats.stats[GFS2_LKS_SRTTVAR]; | ||
274 | __entry->srttb = gl->gl_stats.stats[GFS2_LKS_SRTTB]; | ||
275 | __entry->srttvarb = gl->gl_stats.stats[GFS2_LKS_SRTTVARB]; | ||
276 | __entry->sirt = gl->gl_stats.stats[GFS2_LKS_SIRT]; | ||
277 | __entry->sirtvar = gl->gl_stats.stats[GFS2_LKS_SIRTVAR]; | ||
278 | __entry->dcount = gl->gl_stats.stats[GFS2_LKS_DCOUNT]; | ||
279 | __entry->qcount = gl->gl_stats.stats[GFS2_LKS_QCOUNT]; | ||
280 | ), | ||
281 | |||
282 | TP_printk("%u,%u glock %d:%lld status:%d flags:%02x tdiff:%lld srtt:%lld/%lld srttb:%lld/%lld sirt:%lld/%lld dcnt:%lld qcnt:%lld", | ||
283 | MAJOR(__entry->dev), MINOR(__entry->dev), __entry->gltype, | ||
284 | (unsigned long long)__entry->glnum, | ||
285 | __entry->status, __entry->flags, | ||
286 | (long long)__entry->tdiff, | ||
287 | (long long)__entry->srtt, | ||
288 | (long long)__entry->srttvar, | ||
289 | (long long)__entry->srttb, | ||
290 | (long long)__entry->srttvarb, | ||
291 | (long long)__entry->sirt, | ||
292 | (long long)__entry->sirtvar, | ||
293 | (long long)__entry->dcount, | ||
294 | (long long)__entry->qcount) | ||
295 | ); | ||
296 | |||
239 | /* Section 2 - Log/journal | 297 | /* Section 2 - Log/journal |
240 | * | 298 | * |
241 | * Objectives: | 299 | * Objectives: |
diff --git a/fs/gfs2/util.c b/fs/gfs2/util.c index 53511291fe36..9e7765e8e7b0 100644 --- a/fs/gfs2/util.c +++ b/fs/gfs2/util.c | |||
@@ -25,6 +25,7 @@ struct kmem_cache *gfs2_inode_cachep __read_mostly; | |||
25 | struct kmem_cache *gfs2_bufdata_cachep __read_mostly; | 25 | struct kmem_cache *gfs2_bufdata_cachep __read_mostly; |
26 | struct kmem_cache *gfs2_rgrpd_cachep __read_mostly; | 26 | struct kmem_cache *gfs2_rgrpd_cachep __read_mostly; |
27 | struct kmem_cache *gfs2_quotad_cachep __read_mostly; | 27 | struct kmem_cache *gfs2_quotad_cachep __read_mostly; |
28 | mempool_t *gfs2_bh_pool __read_mostly; | ||
28 | 29 | ||
29 | void gfs2_assert_i(struct gfs2_sbd *sdp) | 30 | void gfs2_assert_i(struct gfs2_sbd *sdp) |
30 | { | 31 | { |
diff --git a/fs/gfs2/util.h b/fs/gfs2/util.h index b432e04600de..a4ce76c67dbb 100644 --- a/fs/gfs2/util.h +++ b/fs/gfs2/util.h | |||
@@ -10,6 +10,8 @@ | |||
10 | #ifndef __UTIL_DOT_H__ | 10 | #ifndef __UTIL_DOT_H__ |
11 | #define __UTIL_DOT_H__ | 11 | #define __UTIL_DOT_H__ |
12 | 12 | ||
13 | #include <linux/mempool.h> | ||
14 | |||
13 | #include "incore.h" | 15 | #include "incore.h" |
14 | 16 | ||
15 | #define fs_printk(level, fs, fmt, arg...) \ | 17 | #define fs_printk(level, fs, fmt, arg...) \ |
@@ -150,6 +152,7 @@ extern struct kmem_cache *gfs2_inode_cachep; | |||
150 | extern struct kmem_cache *gfs2_bufdata_cachep; | 152 | extern struct kmem_cache *gfs2_bufdata_cachep; |
151 | extern struct kmem_cache *gfs2_rgrpd_cachep; | 153 | extern struct kmem_cache *gfs2_rgrpd_cachep; |
152 | extern struct kmem_cache *gfs2_quotad_cachep; | 154 | extern struct kmem_cache *gfs2_quotad_cachep; |
155 | extern mempool_t *gfs2_bh_pool; | ||
153 | 156 | ||
154 | static inline unsigned int gfs2_tune_get_i(struct gfs2_tune *gt, | 157 | static inline unsigned int gfs2_tune_get_i(struct gfs2_tune *gt, |
155 | unsigned int *p) | 158 | unsigned int *p) |
diff --git a/fs/gfs2/xattr.c b/fs/gfs2/xattr.c index e9636591b5d5..2e5ba425cae7 100644 --- a/fs/gfs2/xattr.c +++ b/fs/gfs2/xattr.c | |||
@@ -251,7 +251,7 @@ static int ea_dealloc_unstuffed(struct gfs2_inode *ip, struct buffer_head *bh, | |||
251 | if (!blks) | 251 | if (!blks) |
252 | return 0; | 252 | return 0; |
253 | 253 | ||
254 | rgd = gfs2_blk2rgrpd(sdp, bn); | 254 | rgd = gfs2_blk2rgrpd(sdp, bn, 1); |
255 | if (!rgd) { | 255 | if (!rgd) { |
256 | gfs2_consist_inode(ip); | 256 | gfs2_consist_inode(ip); |
257 | return -EIO; | 257 | return -EIO; |
@@ -1439,7 +1439,7 @@ static int ea_dealloc_block(struct gfs2_inode *ip) | |||
1439 | struct gfs2_holder gh; | 1439 | struct gfs2_holder gh; |
1440 | int error; | 1440 | int error; |
1441 | 1441 | ||
1442 | rgd = gfs2_blk2rgrpd(sdp, ip->i_eattr); | 1442 | rgd = gfs2_blk2rgrpd(sdp, ip->i_eattr, 1); |
1443 | if (!rgd) { | 1443 | if (!rgd) { |
1444 | gfs2_consist_inode(ip); | 1444 | gfs2_consist_inode(ip); |
1445 | return -EIO; | 1445 | return -EIO; |
diff --git a/include/linux/gfs2_ondisk.h b/include/linux/gfs2_ondisk.h index b148087f49a6..fa98bdb073b9 100644 --- a/include/linux/gfs2_ondisk.h +++ b/include/linux/gfs2_ondisk.h | |||
@@ -168,6 +168,7 @@ struct gfs2_rindex { | |||
168 | #define GFS2_RGF_METAONLY 0x00000002 | 168 | #define GFS2_RGF_METAONLY 0x00000002 |
169 | #define GFS2_RGF_DATAONLY 0x00000004 | 169 | #define GFS2_RGF_DATAONLY 0x00000004 |
170 | #define GFS2_RGF_NOALLOC 0x00000008 | 170 | #define GFS2_RGF_NOALLOC 0x00000008 |
171 | #define GFS2_RGF_TRIMMED 0x00000010 | ||
171 | 172 | ||
172 | struct gfs2_rgrp { | 173 | struct gfs2_rgrp { |
173 | struct gfs2_meta_header rg_header; | 174 | struct gfs2_meta_header rg_header; |