Hi all,
On our busy mailserver we see clamd leaking mem like hell. I've now used
valgrind to debug this a bit further.
We use ClamAV devel-20050511 on FreeBSD 5.4 release, but I have seen the
memleak on the devel port since ever. After scanning 10'000 mails, clamd has
grown over 100MB, after
and I did run it a second time (waiting 20 seconds longer)
I don't understand this statement. You waited 20 seconds for what?
I did run clamd (valgrind) 20 seconds longer (about 2 minutes) and
as you can see it consumes a lot more mem.
Martin
___
On Wed, 2005-05-11 at 11:13 +0200, Martin Blapp wrote:
and I did run it a second time (waiting 20 seconds longer)
I don't understand this statement. You waited 20 seconds for what?
I did run clamd (valgrind) 20 seconds longer (about 2 minutes) and
as you can see it consumes a lot more
Hi,
I did run clamd (valgrind) 20 seconds longer (about 2 minutes) and
as you can see it consumes a lot more mem.
How did you request clamd to exit?
kill -4 pidofvalgrind ... CTRL D did not help, also a normal kill didn't. Have I
done something wrong ?
Martin
On Wed, 2005-05-11 at 11:18 +0200, Martin Blapp wrote:
Hi,
I did run clamd (valgrind) 20 seconds longer (about 2 minutes) and
as you can see it consumes a lot more mem.
How did you request clamd to exit?
kill -4 pidofvalgrind ... CTRL D did not help, also a normal kill didn't.
Have
Is this a bug or not ?
--- libclamav/message.cSun Apr 24 12:01:15 2005
+++ libclamav/message.cWed May 11
11:37:57 2005
@@ -850,9 +850,10 @@
m-body_last = m-body_first = (text *)cli_malloc(sizeof(text));
else {
assert(m-body_last != NULL);
-
ok, seems only to be a style problem. Of course comments
are stripped away by the preprozessor.
--- libclamav/message.cSun Apr 24 12:01:15 2005
+++ libclamav/message.cWed May 11 11:37:57 2005
@@ -850,9 +850,10 @@
m-body_last = m-body_first = (text
On Wed, 2005-05-11 at 12:11 +0200, Martin Blapp wrote:
and then clamd/valgrind gets inresponsive. Any other thing I
need to do ?
Probably need to wait for valgrind to finish accounting for the memory.
-trog
signature.asc
Description: This is a digitally signed message part
Hi,
Probably need to wait for valgrind to finish accounting for the memory.
Seems not to be the case. Valgrind doesn't do anything after that, maybe
waiting in a deadlock condition. I've now waited 10 minutes, truss -p PID
doesn't show any activity.
If I use 'kill -4 PID' I get:
--6424-- adding
On Wednesday 11 May 2005 12:03, Martin Blapp wrote:
Hi all,
On our busy mailserver we see clamd leaking mem like hell. I've now used
valgrind to debug this a bit further.
We use ClamAV devel-20050511 on FreeBSD 5.4 release, but I have seen the
memleak on the devel port since ever
The bug is: most likely there are text* which are not treated with
textDestroy().
valgrind -v --tool=memcheck --leak-check=yes --show-reachable=yes
--trace-children=yes --num-callers=9 /docsis/sbin/clamd
+ --leak-resolution=high?
I've tried --leak-resolution=med ... but I'll now use
On Wed, 2005-05-11 at 14:03 +0200, Martin Blapp wrote:
Seems not to be the case. Valgrind doesn't do anything after that, maybe
waiting in a deadlock condition. I've now waited 10 minutes, truss -p PID
doesn't show any activity.
If I wait 10 minutes for clamd finishing its activity and
Hi,
Just as an FYI, I had upgraded to the CVS version (Changelog shows Wed May 4
20:15:33 BST 2005 (njh)) to work around the password protected zip file
problem.
The next morning, clamd was using 638 megs of memory on the system. I killed
it and restarted and 2 hours later it was using 350
I wonder if I could get a clarification of the following message from
the -users mailing list? Nigel seems to be saying that telling clamav-milter
to use clamd via --external doesn't use IPC and is faster than using
the milter's built-in scanner. I would guess the reverse to be true.
From:
I'm about to try turning --external back on to see if it fixes my problem
with clamav-milter calling messages clean while clamscan says the same
messages contain Bagz.something.
By the way, this didn't fix it. I'm still getting clamav-milter saying
messages are clean while clamscan (correctly)
You haven't notified the clamd that is running about your latest virus
database.
I just started clamd. Are you saying it needs to be notified right
after it starts?
---
Jef
Jef Poskanzer [EMAIL PROTECTED] http://www.acme.com/jef/
___
OS: Linux 2.4.26
I just tried to upgrade from 0.83 to 0.85 had problems with not being
able to open /dev/console. 0.83 did something like this:
1830 if((cpt = cfgopt(copt, LogFile))) {
snip
1841 } else
1842 logFile = /dev/console;
1843
I have found the source of the issue I reported to clamav-users last
week (originally with a subject of Input/Output error with .zip file
under 0.84), with the help of a pointer from Andrey Melnikoff.
The issue is: If an unencrypted .zip file over 32KB in size is
compressed using the PKZIP
18 matches
Mail list logo