Bug 144446

Summary: [patch] db(3) fails with large block sizes
Product: Base System Reporter: Peter Jeremy <peterjeremy>
Component: binAssignee: Andriy Gapon <avg>
Status: Closed FIXED    
Severity: Affects Only Me    
Priority: Normal    
Version: 8.0-STABLE   
Hardware: Any   
OS: Any   
Attachments:
Description Flags
file.diff
none
file.diff
none
file.diff none

Description Peter Jeremy 2010-03-03 11:30:01 UTC
	Whilst trying to port db(3) to a Solaris system, I have identified
	two issues with the existing hash(3) code.

	Firstly, when creating a new hash database, the bucket size
	defaults to the st_blksize of the file (hash/hash.c::init_hash()).
	There is no sanity checking to ensure that st_blksize is within
	valid limits (hash/hash.h defined MAX_BSIZE as 65536).

	In FreeBSD, st_blksize is currently hardwired to PAGE_SIZE in
	kern/vfs_vnops.c::vn_stat() so this is purely a theoretical
	issue on FreeBSD.  Solaris exposes the blocksize from the
	underlying filesystem - and in the case of ZFS, this is 128KB,
	which exceeds MAX_BSIZE.  In my case, the symptoms were that
	when sequentially reading the database (via DB->seq()), the
	returned keys were padded with 64KB of NULs.

	Secondly, when the bucket size is set to 64KB (MAX_BSIZE),
	non-trivial databases crash reporting:
	    "HASH: Out of overflow pages.  Increase page size"
	It's not clear what triggers this.

	Thirdly, whilst writing this PR, I've noticed that hash(3)
	states that the default hash table bucket size is 256 bytes.
	The actual default (as per DEF_BUCKET_SIZE in hash/hash.h)
	is 4096 bytes.

Fix: First patch adds check for excessive st_blksize.  Note that this
	assumes that st_blksize is a power of 2.  It might be reasonable
	to verify that st_blksize isn't too small but I'm not sure what
	the lower limit is.

This is simply a work-around for the second problem.  I'm not
	certain what the actual problem is bu this avoids it.

The following updates the man page to contain the current
	default bucket size as per DEF_BUCKET_SIZE in hash/hash.h
How-To-Repeat: 	First problem can't be reproduced on FreeBSD but occurs on
	at least Solaris and OpenSolaris with ZFS.

	The second problem can be reproduced by extending the db
	test tool (lib/libc/db/test/run.test, test20) to include
	a bucket size of 65536.  Based on the progression of
	fill factor's in test20, the logical fill factors are
	2735 3647 5471, however the test fails with each of these
	values as well as 8, 341 and 10001.  All other bucket
	sizes appear to work successfully - which suggests there
	is a problem with this particular bucket size.
Comment 1 dfilter service freebsd_committer freebsd_triage 2010-04-05 11:02:06 UTC
Author: avg
Date: Mon Apr  5 10:01:53 2010
New Revision: 206177
URL: http://svn.freebsd.org/changeset/base/206177

Log:
  hash.3: fix a factual mistake in the man page
  
  PR:		bin/144446
  Submitted by:	Peter Jeremy <peterjeremy@acm.org>
  MFC after:	3 days

Modified:
  head/lib/libc/db/man/hash.3

Modified: head/lib/libc/db/man/hash.3
==============================================================================
--- head/lib/libc/db/man/hash.3	Mon Apr  5 09:26:03 2010	(r206176)
+++ head/lib/libc/db/man/hash.3	Mon Apr  5 10:01:53 2010	(r206177)
@@ -78,7 +78,7 @@ The
 element
 defines the
 .Nm
-table bucket size, and is, by default, 256 bytes.
+table bucket size, and is, by default, 4096 bytes.
 It may be preferable to increase the page size for disk-resident tables
 and tables with large data items.
 .It Va ffactor
_______________________________________________
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscribe@freebsd.org"
Comment 2 dfilter service freebsd_committer freebsd_triage 2010-04-05 11:12:35 UTC
Author: avg
Date: Mon Apr  5 10:12:21 2010
New Revision: 206178
URL: http://svn.freebsd.org/changeset/base/206178

Log:
  libc/db/hash: cap auto-tuned block size with a value that actually works
  
  This fix mostly matters after r206129 that made it possible for
  st_blksize to be greater than 4K.  For this reason, this change should
  be MFC-ed before r206129.
  Also, it seems that all FreeBSD uitlities that use db(3) hash databases
  and create new databases in files, specify their own block size value
  and thus do not depend on block size autotuning.
  
  PR:		bin/144446
  Submitted by:	Peter Jeremy <peterjeremy@acm.org>
  MFC after:	5 days

Modified:
  head/lib/libc/db/hash/hash.c
  head/lib/libc/db/hash/hash.h

Modified: head/lib/libc/db/hash/hash.c
==============================================================================
--- head/lib/libc/db/hash/hash.c	Mon Apr  5 10:01:53 2010	(r206177)
+++ head/lib/libc/db/hash/hash.c	Mon Apr  5 10:12:21 2010	(r206178)
@@ -293,6 +293,8 @@ init_hash(HTAB *hashp, const char *file,
 		if (stat(file, &statbuf))
 			return (NULL);
 		hashp->BSIZE = statbuf.st_blksize;
+		if (hashp->BSIZE > MAX_BSIZE)
+			hashp->BSIZE = MAX_BSIZE;
 		hashp->BSHIFT = __log2(hashp->BSIZE);
 	}
 

Modified: head/lib/libc/db/hash/hash.h
==============================================================================
--- head/lib/libc/db/hash/hash.h	Mon Apr  5 10:01:53 2010	(r206177)
+++ head/lib/libc/db/hash/hash.h	Mon Apr  5 10:12:21 2010	(r206178)
@@ -118,7 +118,7 @@ typedef struct htab	 {		/* Memory reside
 /*
  * Constants
  */
-#define	MAX_BSIZE		65536		/* 2^16 */
+#define	MAX_BSIZE		32768		/* 2^15 but should be 65536 */
 #define MIN_BUFFERS		6
 #define MINHDRSIZE		512
 #define DEF_BUFSIZE		65536		/* 64 K */
_______________________________________________
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscribe@freebsd.org"
Comment 3 Peter Jeremy 2010-04-06 01:09:04 UTC
Following r206129, FreeBSD no longer has a hardwired st_blksize
so the first problem should now be reproducable on FreeBSD.

The scenario where I found the bug was the perl 5.8 DB_File test
db-hash.t on a ZFS filesystem.  The problem should therefore be
reproducable on 9-current with r206129 as follows:

1) Set WRKDIRPREFIX to a ZFS filesystem.
2) cd /usr/ports/lang/perl5.8
3) make
4) cd $WRKDIRPREFIX/usr/ports/lang/perl5.8/work/perl-5.8.9/t
5) ../perl ../ext/DB_File/t/db-hash.t
   (it's possible perl will need to be installed for this to run)
This should result in a core dump following test 21.

-- 
Peter Jeremy
Comment 4 Volker Werth freebsd_committer freebsd_triage 2010-08-12 21:46:11 UTC
State Changed
From-To: open->patched

already committed and MFC'ed 


Comment 5 Volker Werth freebsd_committer freebsd_triage 2010-08-12 21:46:11 UTC
Responsible Changed
From-To: freebsd-bugs->avg

Andriy, it seems to be already MFC'ed. Can this be closed now?
Comment 6 Andriy Gapon freebsd_committer freebsd_triage 2010-08-21 17:09:23 UTC
State Changed
From-To: patched->closed

Fixed.