Re: fsck_ffs diff needs testing
Please remove this diff from your tree for the moment. I have found a bug. More to follow soon. -Otto
fsck_ffs diff needs testing
Hi,
I got little feedeback on this diff posed in a rather long thread, so
I am posting it again.
Please test this, it makes fsck_ffs much faster (especially with -p)
and less memory hungry in a lot of cases.
Note that to force a check with -p you need to unmount the filesystem,
mosty practical is do it in single user mode:
# shutdown now
..
# umount -a
# fsck -pf
(Don't forget the unmount!)
Other way: pull the plug on a machine and let it reboot. That is a
more realistic check and it is safe because you already backup up all
your data, right?
-Otto
Index: dir.c
===
RCS file: /cvs/src/sbin/fsck_ffs/dir.c,v
retrieving revision 1.24
diff -u -p -r1.24 dir.c
--- dir.c 27 Oct 2009 23:59:32 - 1.24
+++ dir.c 4 Apr 2011 09:15:49 -
@@ -443,8 +443,8 @@ linkup(ino_t orphan, ino_t parentdir)
idesc.id_type = ADDR;
idesc.id_func = pass4check;
idesc.id_number = oldlfdir;
- adjust(idesc, lncntp[oldlfdir] + 1);
- lncntp[oldlfdir] = 0;
+ adjust(idesc, ILNCOUNT(oldlfdir) + 1);
+ ILNCOUNT(oldlfdir) = 0;
dp = ginode(lfdir);
}
if (GET_ISTATE(lfdir) != DFOUND) {
@@ -457,7 +457,7 @@ linkup(ino_t orphan, ino_t parentdir)
printf(\n\n);
return (0);
}
- lncntp[orphan]--;
+ ILNCOUNT(orphan)--;
if (lostdir) {
if ((changeino(orphan, .., lfdir) ALTERED) == 0
parentdir != (ino_t)-1)
@@ -465,7 +465,7 @@ linkup(ino_t orphan, ino_t parentdir)
dp = ginode(lfdir);
DIP_SET(dp, di_nlink, DIP(dp, di_nlink) + 1);
inodirty();
- lncntp[lfdir]++;
+ ILNCOUNT(lfdir)++;
pwarn(DIR I=%u CONNECTED. , orphan);
if (parentdir != (ino_t)-1) {
printf(PARENT WAS I=%u\n, parentdir);
@@ -476,7 +476,7 @@ linkup(ino_t orphan, ino_t parentdir)
* fixes the parent link count so that fsck does
* not need to be rerun.
*/
- lncntp[parentdir]++;
+ ILNCOUNT(parentdir)++;
}
if (preen == 0)
printf(\n);
@@ -636,7 +636,7 @@ allocdir(ino_t parent, ino_t request, in
DIP_SET(dp, di_nlink, 2);
inodirty();
if (ino == ROOTINO) {
- lncntp[ino] = DIP(dp, di_nlink);
+ ILNCOUNT(ino) = DIP(dp, di_nlink);
cacheino(dp, ino);
return(ino);
}
@@ -650,8 +650,8 @@ allocdir(ino_t parent, ino_t request, in
inp-i_dotdot = parent;
SET_ISTATE(ino, GET_ISTATE(parent));
if (GET_ISTATE(ino) == DSTATE) {
- lncntp[ino] = DIP(dp, di_nlink);
- lncntp[parent]++;
+ ILNCOUNT(ino) = DIP(dp, di_nlink);
+ ILNCOUNT(parent)++;
}
dp = ginode(parent);
DIP_SET(dp, di_nlink, DIP(dp, di_nlink) + 1);
Index: extern.h
===
RCS file: /cvs/src/sbin/fsck_ffs/extern.h,v
retrieving revision 1.10
diff -u -p -r1.10 extern.h
--- extern.h25 Jun 2007 19:59:55 - 1.10
+++ extern.h4 Apr 2011 09:15:49 -
@@ -54,6 +54,7 @@ int ftypeok(union dinode *);
void getpathname(char *, size_t, ino_t, ino_t);
void inocleanup(void);
void inodirty(void);
+struct inostat *inoinfo(ino_t);
intlinkup(ino_t, ino_t);
intmakeentry(ino_t, ino_t, char *);
void pass1(void);
Index: fsck.h
===
RCS file: /cvs/src/sbin/fsck_ffs/fsck.h,v
retrieving revision 1.23
diff -u -p -r1.23 fsck.h
--- fsck.h 10 Jun 2008 23:10:29 - 1.23
+++ fsck.h 4 Apr 2011 09:15:49 -
@@ -66,6 +66,19 @@ union dinode {
#define BUFSIZ 1024
#endif
+/*
+ * Each inode on the file system is described by the following structure.
+ * The linkcnt is initially set to the value in the inode. Each time it
+ * is found during the descent in passes 2, 3, and 4 the count is
+ * decremented. Any inodes whose count is non-zero after pass 4 needs to
+ * have its link count adjusted by the value remaining in ino_linkcnt.
+ */
+struct inostat {
+ charino_state; /* state of inode, see below */
+ charino_type; /* type of inode */
+ short ino_linkcnt;/* number of links not found */
+};
+
#defineUSTATE 01 /* inode not allocated */
#defineFSTATE 02 /* inode is file */
#defineDSTATE 03 /* inode is directory */
@@ -73,12 +86,20 @@ union dinode {
#defineDCLEAR 05 /* directory is to be cleared */
#defineFCLEAR 06 /* file is to be
Re: fsck_ffs diff needs testing
On Thu, Apr 07, 2011 at 11:08:05AM +0200, Otto Moerbeek wrote:
Hi,
I got little feedeback on this diff posed in a rather long thread, so
I am posting it again.
Please test this, it makes fsck_ffs much faster (especially with -p)
and less memory hungry in a lot of cases.
Note that to force a check with -p you need to unmount the filesystem,
mosty practical is do it in single user mode:
# shutdown now
..
# umount -a
# fsck -pf
(Don't forget the unmount!)
Other way: pull the plug on a machine and let it reboot. That is a
more realistic check and it is safe because you already backup up all
your data, right?
I went the unplug route on the i386 box and amd64 box I have here, and
the diff seems to work. Reads ok. I say go for it.
Ken
-Otto
Index: dir.c
===
RCS file: /cvs/src/sbin/fsck_ffs/dir.c,v
retrieving revision 1.24
diff -u -p -r1.24 dir.c
--- dir.c 27 Oct 2009 23:59:32 - 1.24
+++ dir.c 4 Apr 2011 09:15:49 -
@@ -443,8 +443,8 @@ linkup(ino_t orphan, ino_t parentdir)
idesc.id_type = ADDR;
idesc.id_func = pass4check;
idesc.id_number = oldlfdir;
- adjust(idesc, lncntp[oldlfdir] + 1);
- lncntp[oldlfdir] = 0;
+ adjust(idesc, ILNCOUNT(oldlfdir) + 1);
+ ILNCOUNT(oldlfdir) = 0;
dp = ginode(lfdir);
}
if (GET_ISTATE(lfdir) != DFOUND) {
@@ -457,7 +457,7 @@ linkup(ino_t orphan, ino_t parentdir)
printf(\n\n);
return (0);
}
- lncntp[orphan]--;
+ ILNCOUNT(orphan)--;
if (lostdir) {
if ((changeino(orphan, .., lfdir) ALTERED) == 0
parentdir != (ino_t)-1)
@@ -465,7 +465,7 @@ linkup(ino_t orphan, ino_t parentdir)
dp = ginode(lfdir);
DIP_SET(dp, di_nlink, DIP(dp, di_nlink) + 1);
inodirty();
- lncntp[lfdir]++;
+ ILNCOUNT(lfdir)++;
pwarn(DIR I=%u CONNECTED. , orphan);
if (parentdir != (ino_t)-1) {
printf(PARENT WAS I=%u\n, parentdir);
@@ -476,7 +476,7 @@ linkup(ino_t orphan, ino_t parentdir)
* fixes the parent link count so that fsck does
* not need to be rerun.
*/
- lncntp[parentdir]++;
+ ILNCOUNT(parentdir)++;
}
if (preen == 0)
printf(\n);
@@ -636,7 +636,7 @@ allocdir(ino_t parent, ino_t request, in
DIP_SET(dp, di_nlink, 2);
inodirty();
if (ino == ROOTINO) {
- lncntp[ino] = DIP(dp, di_nlink);
+ ILNCOUNT(ino) = DIP(dp, di_nlink);
cacheino(dp, ino);
return(ino);
}
@@ -650,8 +650,8 @@ allocdir(ino_t parent, ino_t request, in
inp-i_dotdot = parent;
SET_ISTATE(ino, GET_ISTATE(parent));
if (GET_ISTATE(ino) == DSTATE) {
- lncntp[ino] = DIP(dp, di_nlink);
- lncntp[parent]++;
+ ILNCOUNT(ino) = DIP(dp, di_nlink);
+ ILNCOUNT(parent)++;
}
dp = ginode(parent);
DIP_SET(dp, di_nlink, DIP(dp, di_nlink) + 1);
Index: extern.h
===
RCS file: /cvs/src/sbin/fsck_ffs/extern.h,v
retrieving revision 1.10
diff -u -p -r1.10 extern.h
--- extern.h 25 Jun 2007 19:59:55 - 1.10
+++ extern.h 4 Apr 2011 09:15:49 -
@@ -54,6 +54,7 @@ int ftypeok(union dinode *);
void getpathname(char *, size_t, ino_t, ino_t);
void inocleanup(void);
void inodirty(void);
+struct inostat *inoinfo(ino_t);
int linkup(ino_t, ino_t);
int makeentry(ino_t, ino_t, char *);
void pass1(void);
Index: fsck.h
===
RCS file: /cvs/src/sbin/fsck_ffs/fsck.h,v
retrieving revision 1.23
diff -u -p -r1.23 fsck.h
--- fsck.h10 Jun 2008 23:10:29 - 1.23
+++ fsck.h4 Apr 2011 09:15:49 -
@@ -66,6 +66,19 @@ union dinode {
#define BUFSIZ 1024
#endif
+/*
+ * Each inode on the file system is described by the following structure.
+ * The linkcnt is initially set to the value in the inode. Each time it
+ * is found during the descent in passes 2, 3, and 4 the count is
+ * decremented. Any inodes whose count is non-zero after pass 4 needs to
+ * have its link count adjusted by the value remaining in ino_linkcnt.
+ */
+struct inostat {
+ charino_state; /* state of inode, see below */
+ charino_type; /* type of inode */
+ short ino_linkcnt;/* number of links not found */
+};
+
#define USTATE 01 /* inode not allocated */
#define FSTATE 02 /* inode is file */
#define DSTATE 03 /* inode is
Re: fsck_ffs diff needs testing
On 2011-04-07 11.08, Otto Moerbeek wrote:
Hi,
I got little feedeback on this diff posed in a rather long thread, so
I am posting it again.
Please test this, it makes fsck_ffs much faster (especially with -p)
and less memory hungry in a lot of cases.
I've run it on a variety of file systems now and first of all, it seems
to work without regressions. It also looks like it has about 2/3 of its
previous memory footprint. However, I see very little increase in speed:
Unpatched, run with time /sbin/fsck_ffs -pf /dev/rraid0[adefghi]:
/dev/rraid0a: 2874 files, 82005 used, 46810 free (866 frags, 5743
blocks, 0.7% fragmentation)
/dev/rraid0d: 1477 files, 144486 used, 886745 free (553 frags, 110774
blocks, 0.1% fragmentation)
/dev/rraid0e: 224749 files, 1624996 used, 438155 free (48771 frags,
48673 blocks, 2.4% fragmentation)
/dev/rraid0f: 9437 files, 103157 used, 4024794 free (3370 frags, 502678
blocks, 0.1% fragmentation)
/dev/rraid0g: 498655 files, 6190187 used, 4128604 free (17188 frags,
513927 blocks, 0.2% fragmentation)
/dev/rraid0h: 90459 files, 445770 used, 2649301 free (20357 frags,
328618 blocks, 0.7% fragmentation)
/dev/rraid0i: 35896 files, 666529 used, 1266782 free (9006 frags, 157222
blocks, 0.5% fragmentation)
1m20.35s real 0m2.57s user 0m4.29s system
Patched, run with time /root/fsck_ffs -pf /dev/rraid0[adefghi]:
/dev/rraid0a: 2874 files, 82005 used, 46810 free (866 frags, 5743
blocks, 0.7% fragmentation)
/dev/rraid0d: 1477 files, 144486 used, 886745 free (553 frags, 110774
blocks, 0.1% fragmentation)
/dev/rraid0e: 224749 files, 1624996 used, 438155 free (48771 frags,
48673 blocks, 2.4% fragmentation)
/dev/rraid0f: 9437 files, 103157 used, 4024794 free (3370 frags, 502678
blocks, 0.1% fragmentation)
/dev/rraid0g: 498655 files, 6190187 used, 4128604 free (17188 frags,
513927 blocks, 0.2% fragmentation)
/dev/rraid0h: 90459 files, 445770 used, 2649301 free (20357 frags,
328618 blocks, 0.7% fragmentation)
/dev/rraid0i: 35896 files, 666529 used, 1266782 free (9006 frags, 157222
blocks, 0.5% fragmentation)
1m18.52s real 0m1.32s user 0m3.85s system
The file systems are organized like this:
skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0a
# newfs command for /dev/rraid0a
newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s
525856 /dev/rraid0a
skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0d
# newfs command for /dev/rraid0d
newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s
4197120 /dev/rraid0d
skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0e
# newfs command for /dev/rraid0e
newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s
8390400 /dev/rraid0e
skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0f
# newfs command for /dev/rraid0f
newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s
16780800 /dev/rraid0f
skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0g
# newfs command for /dev/rraid0g
newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s
41944320 /dev/rraid0g
skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0h
# newfs command for /dev/rraid0h
newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s
12583680 /dev/rraid0h
skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0i
# newfs command for /dev/rraid0i
newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s
7857920 /dev/rraid0i
skynet:~/fsck_ffs_patch#
Btw, I noticed a small change in this diff compared to the one you posted
in the previous discussion thread. I assume this is the correct one to use?
skynet:~# diff fsck_ffs.patch fsck_ffs_2.patch
158c158
+++ inode.c 31 Mar 2011 16:34:27 -
---
+++ inode.c 4 Apr 2011 09:16:36 -
203c203
+ info[i].ino_state = USTATE;
---
+ SET_ISTATE(i, USTATE);
Regards,
/Benny
Note that to force a check with -p you need to unmount the filesystem,
mosty practical is do it in single user mode:
# shutdown now
..
# umount -a
# fsck -pf
(Don't forget the unmount!)
Other way: pull the plug on a machine and let it reboot. That is a
more realistic check and it is safe because you already backup up all
your data, right?
-Otto
Index: dir.c
===
RCS file: /cvs/src/sbin/fsck_ffs/dir.c,v
retrieving revision 1.24
diff -u -p -r1.24 dir.c
--- dir.c 27 Oct 2009 23:59:32 - 1.24
+++ dir.c 4 Apr 2011 09:15:49 -
@@ -443,8 +443,8 @@ linkup(ino_t orphan, ino_t parentdir)
idesc.id_type = ADDR;
idesc.id_func = pass4check;
idesc.id_number = oldlfdir;
- adjust(idesc, lncntp[oldlfdir] + 1);
- lncntp[oldlfdir] = 0;
+ adjust(idesc, ILNCOUNT(oldlfdir) + 1);
+ ILNCOUNT(oldlfdir) = 0;
dp = ginode(lfdir);
}
if (GET_ISTATE(lfdir) != DFOUND) {
@@ -457,7 +457,7 @@ linkup(ino_t orphan, ino_t parentdir)
printf(\n\n);
Re: fsck_ffs diff needs testing
On Thu, Apr 07, 2011 at 04:28:25PM +0200, Benny Lofgren wrote: On 2011-04-07 15.06, Otto Moerbeek wrote: On Thu, Apr 07, 2011 at 02:34:27PM +0200, Benny Lofgren wrote: On 2011-04-07 11.08, Otto Moerbeek wrote: Hi, I got little feedeback on this diff posed in a rather long thread, so I am posting it again. Please test this, it makes fsck_ffs much faster (especially with -p) and less memory hungry in a lot of cases. I've run it on a variety of file systems now and first of all, it seems to work without regressions. It also looks like it has about 2/3 of its previous memory footprint. However, I see very little increase in speed: Were your filesystems mounted with softdep? Also, the most gain is realized with ffs1 filesystems. Apart from the / file system they had been mounted with softdep, yes. However they were of course unmounted when I ran fsck_ffs. Does softdep actually alter the physical file system structure on disk? I have always been under the impression that softdep affects stuff like how things are buffered and the order of which things are written to disk rather than *what* is written, am I mistaken? softdep does not change the layout. But only filesystems which were mounted with softdep get this optimization. There's a flag in teh superblock to signal that. Filesystem mounted with softdep have better guarantees about the cylinder group headers being consistent. -Otto And yes, they were all ffs1 filesystems. I have a 13 TB ffs2 file system on that machine as well, didn't try that now because I wanted to fire off a quick reply to your request for testing, and also you mentioned that ffs2 is less (if at all) affected by this patch. I'll be glad to run a test against that volume as well, if you think there's an interest. Expect about 50 minutes a run for one pass with the old and one pass with the new code though. :-) Regards, /Benny Unpatched, run with time /sbin/fsck_ffs -pf /dev/rraid0[adefghi]: [snip] 1m20.35s real 0m2.57s user 0m4.29s system Patched, run with time /root/fsck_ffs -pf /dev/rraid0[adefghi]: [snip] 1m18.52s real 0m1.32s user 0m3.85s system The file systems are organized like this: skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0a # newfs command for /dev/rraid0a newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s 525856 /dev/rraid0a skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0d # newfs command for /dev/rraid0d newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s 4197120 /dev/rraid0d skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0e # newfs command for /dev/rraid0e newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s 8390400 /dev/rraid0e skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0f # newfs command for /dev/rraid0f newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s 16780800 /dev/rraid0f skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0g # newfs command for /dev/rraid0g newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s 41944320 /dev/rraid0g skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0h # newfs command for /dev/rraid0h newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s 12583680 /dev/rraid0h skynet:~/fsck_ffs_patch# dumpfs -m /dev/rraid0i # newfs command for /dev/rraid0i newfs -O 1 -b 16384 -e 4096 -f 2048 -g 16384 -h 64 -m 5 -o time -s 7857920 /dev/rraid0i skynet:~/fsck_ffs_patch# Btw, I noticed a small change in this diff compared to the one you posted in the previous discussion thread. I assume this is the correct one to use? Yes, though that's the most recent version. -Otto skynet:~# diff fsck_ffs.patch fsck_ffs_2.patch 158c158 +++ inode.c 31 Mar 2011 16:34:27 - --- +++ inode.c 4 Apr 2011 09:16:36 - 203c203 + info[i].ino_state = USTATE; --- + SET_ISTATE(i, USTATE); Regards, /Benny -- internetlabbet.se / work: +46 8 551 124 80 / Words must Benny L?fgren/ mobile: +46 70 718 11 90 / be weighed, / fax:+46 8 551 124 89/not counted. /email: benny -at- internetlabbet.se
Re: fsck_ffs diff needs testing
On Thu, Apr 07, 2011 at 11:05:42AM -0500, Amit Kulkarni wrote: Also, depending on the usage patterns, you might have a fs where high numbered inodes are used, while the fs itself is pretty empty. Filling up a fs with lots of files and them removing a lot of them is an example that could lead to such a situation. This diff does not speed things up in such cases. ...might have an impact in my case, since I often do things like rebuilding the system including tons of packages on this machine, and that use case of course perfectly matches what you say above. I think I'll remake these file systems and run the test again just to satisfy my curiosity. But that'll have to wait until after dinner. :-) Anyway, I see improvements both in memory usage and in speed, and I see no obvoius malfunctions, so I'd say it's a go. Hi Otto, Comparing your diff with FreeBSD svn (not cvs, they dropped cvs! my bad on the initial comment) after Benny pointed this out. http://svn.freebsd.org/viewvc/base/head/sbin/fsck_ffs/pass1.c?revision=201708view=markup Look at this comment inside the file /* * This optimization speeds up future runs of fsck * by trimming down the number of inodes in cylinder * groups that formerly had many inodes but now have * fewer in use. */ and the commit entry by McKusick for rev 188110 Update the actions previously attempted by the -D option to make them robust. With these changes fsck is now able to detect and reliably rebuild corrupted cylinder group maps. The -D option is no longer necessary as it has been replaced by a prompt asking whether the corrupted cylinder group should be rebuilt and doing so when requested. These actions are only offered and taken when running fsck in manual mode. Corrupted cylinder groups found during preen mode cause the fsck to fail. Add the -r option to free up excess unused inodes. Decreasing the number of preallocated inodes reduces the running time of future runs of fsck and frees up space that can allocated to files. The -r option is ignored when running in preen mode. -- Will you please please please integrate that part of the code too!!! This is absolutely useful to have and is a fairly common situation. It will make fsck better and be a good way to partially defrag your fs. Or did you plan to keep that for later after more testing on this diff? Thanks, amit Yes, I go step by step. -Otto
Re: fsck_ffs diff needs testing
Also, depending on the usage patterns, you might have a fs where high numbered inodes are used, while the fs itself is pretty empty. Filling up a fs with lots of files and them removing a lot of them is an example that could lead to such a situation. This diff does not speed things up in such cases. ...might have an impact in my case, since I often do things like rebuilding the system including tons of packages on this machine, and that use case of course perfectly matches what you say above. I think I'll remake these file systems and run the test again just to satisfy my curiosity. But that'll have to wait until after dinner. :-) Anyway, I see improvements both in memory usage and in speed, and I see no obvoius malfunctions, so I'd say it's a go. Hi Otto, Comparing your diff with FreeBSD svn (not cvs, they dropped cvs! my bad on the initial comment) after Benny pointed this out. http://svn.freebsd.org/viewvc/base/head/sbin/fsck_ffs/pass1.c?revision=201708view=markup Look at this comment inside the file /* * This optimization speeds up future runs of fsck * by trimming down the number of inodes in cylinder * groups that formerly had many inodes but now have * fewer in use. */ and the commit entry by McKusick for rev 188110 Update the actions previously attempted by the -D option to make them robust. With these changes fsck is now able to detect and reliably rebuild corrupted cylinder group maps. The -D option is no longer necessary as it has been replaced by a prompt asking whether the corrupted cylinder group should be rebuilt and doing so when requested. These actions are only offered and taken when running fsck in manual mode. Corrupted cylinder groups found during preen mode cause the fsck to fail. Add the -r option to free up excess unused inodes. Decreasing the number of preallocated inodes reduces the running time of future runs of fsck and frees up space that can allocated to files. The -r option is ignored when running in preen mode. -- Will you please please please integrate that part of the code too!!! This is absolutely useful to have and is a fairly common situation. It will make fsck better and be a good way to partially defrag your fs. Or did you plan to keep that for later after more testing on this diff? Thanks, amit
Re: fsck_ffs diff needs testing
Thanks! I understand. On Thu, Apr 7, 2011 at 11:09 AM, Otto Moerbeek o...@drijf.net wrote: On Thu, Apr 07, 2011 at 11:05:42AM -0500, Amit Kulkarni wrote: Also, depending on the usage patterns, you might have a fs where high numbered inodes are used, while the fs itself is pretty empty. Filling up a fs with lots of files and them removing a lot of them is an example that could lead to such a situation. This diff does not speed things up in such cases. ...might have an impact in my case, since I often do things like rebuilding the system including tons of packages on this machine, and that use case of course perfectly matches what you say above. I think I'll remake these file systems and run the test again just to satisfy my curiosity. But that'll have to wait until after dinner. :-) Anyway, I see improvements both in memory usage and in speed, and I see no obvoius malfunctions, so I'd say it's a go. Hi Otto, Comparing your diff with FreeBSD svn (not cvs, they dropped cvs! my bad on the initial comment) after Benny pointed this out. http://svn.freebsd.org/viewvc/base/head/sbin/fsck_ffs/pass1.c?revision=201708 view=markup Look at this comment inside the file /* * This optimization speeds up future runs of fsck * by trimming down the number of inodes in cylinder * groups that formerly had many inodes but now have * fewer in use. */ and the commit entry by McKusick for rev 188110 Update the actions previously attempted by the -D option to make them robust. With these changes fsck is now able to detect and reliably rebuild corrupted cylinder group maps. The -D option is no longer necessary as it has been replaced by a prompt asking whether the corrupted cylinder group should be rebuilt and doing so when requested. These actions are only offered and taken when running fsck in manual mode. Corrupted cylinder groups found during preen mode cause the fsck to fail. Add the -r option to free up excess unused inodes. Decreasing the number of preallocated inodes reduces the running time of future runs of fsck and frees up space that can allocated to files. The -r option is ignored when running in preen mode. -- Will you please please please integrate that part of the code too!!! This is absolutely useful to have and is a fairly common situation. It will make fsck better and be a good way to partially defrag your fs. Or did you plan to keep that for later after more testing on this diff? Thanks, amit Yes, I go step by step. -Otto
