cvs commit: src/sys/vm vm_object.c vm_pageout.c

Pawel Jakub Dawidek pjd at FreeBSD.org
Fri Aug 12 20:01:49 GMT 2005


On Wed, Aug 10, 2005 at 12:17:36AM +0000, Tor Egge wrote:
+> tegge       2005-08-10 00:17:36 UTC
+> 
+>   FreeBSD src repository
+> 
+>   Modified files:
+>     sys/vm               vm_object.c vm_pageout.c 
+>   Log:
+>   Don't allow pagedaemon to skip pages while scanning PQ_ACTIVE or PQ_INACTIVE
+>   due to the vm object being locked.
+>   
+>   When a process writes large amounts of data to a file, the vm object associated
+>   with that file can contain most of the physical pages on the machine.  If the
+>   process is preempted while holding the lock on the vm object, pagedaemon would
+>   be able to move very few pages from PQ_INACTIVE to PQ_CACHE or from PQ_ACTIVE
+>   to PQ_INACTIVE, resulting in unlimited cleaning of dirty pages belonging to
+>   other vm objects.
+>   
+>   Temporarily unlock the page queues lock while locking vm objects to avoid lock
+>   order violation.  Detect and handle relevant page queue changes.
+>   
+>   This change depends on both the lock portion of struct vm_object and normal
+>   struct vm_page being type stable.

I'm not sure this fix the problem completely.

I can still reproduce it with GELI, which is allocating in-kernel memory
for every write request:

	# geli onetime -s 4096 ad0a
	# newfs /dev/ad0a.eli
	# mount /dev/ad0a.eli /mnt
	# dd if=/dev/zero of=/mnt/test bs=1m

In top(1) output I see how memory goes down to few hundred kilo bytes
and after some time I get:

GEOM_ELI: Cannot allocate 136192 bytes.
GEOM_ELI: g_eli_crypto_run() failed (error=12). ad0a.eli[WRITE(offset=4619829248, length=131072)]
GEOM_ELI: Cannot allocate 136192 bytes.
GEOM_ELI: g_eli_crypto_run() failed (error=12). ad0a.eli[WRITE(offset=4620746752, length=131072)]
[...]

There is no memory leak in GELI:

	# vmstat -m | grep eli
	     eli data     2     1K       -     5309  32,256,512,1024

It is much better than it was before your fix, but I think the problem
still exist, it's only harder to run into.

-- 
Pawel Jakub Dawidek                       http://www.wheel.pl
pjd at FreeBSD.org                           http://www.FreeBSD.org
FreeBSD committer                         Am I Evil? Yes, I Am!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/cvs-all/attachments/20050812/7beae260/attachment.bin


More information about the cvs-all mailing list