Backport various fixes for memory leaks, crashes and trivial fixes. http://gitweb.freedesktop.org/?p=xorg/xserver.git;a=commit;h=b5fde366e2e21234ac0b81222fd5c42ca3e49cba Properly free device devPrivates - memory leak fix. author Eamon Walsh <ewalsh@tycho.nsa.gov> Wed, 7 Mar 2007 17:29:55 +0000 (12:29 -0500) committer Eamon Walsh <ewalsh@moss-uranus.epoch.ncsc.mil> Wed, 7 Mar 2007 17:29:55 +0000 (12:29 -0500) commit b5fde366e2e21234ac0b81222fd5c42ca3e49cba tree 21c4d64fc97f8a7b0ec39f7e7a408540dd78691e tree parent a3d2c5d622d9ca36d6fa2966aff09524e3ea39ac commit | commitdiff Properly free device devPrivates - memory leak fix. http://gitweb.freedesktop.org/?p=xorg/xserver.git;a=commit;h=d9e079d2a385203fdd18d958cfc19d759cab4ba8 Zero out client devPrivates on allocation. author Eamon Walsh <ewalsh@tycho.nsa.gov> Fri, 22 Dec 2006 18:07:09 +0000 (13:07 -0500) committer Eamon Walsh <ewalsh@moss-uranus.epoch.ncsc.mil> Fri, 22 Dec 2006 18:07:09 +0000 (13:07 -0500) commit d9e079d2a385203fdd18d958cfc19d759cab4ba8 tree 20548a2b4ab268c5a32c5883f3d06816c9f5d43e tree parent e437f357b6850a6c87ca6696870b3abd40e5b8ed commit | commitdiff Zero out client devPrivates on allocation. http://gitweb.freedesktop.org/?p=xorg/xserver.git;a=commit;h=967865393c08f53c70da9c8cd2824978f2601aa4 Plug memory leak in doLoadModule() author Alan Coopersmith <alan.coopersmith@sun.com> Thu, 25 Jan 2007 04:20:48 +0000 (20:20 -0800) committer Eric Anholt <eric@anholt.net> Tue, 13 Feb 2007 21:25:46 +0000 (13:25 -0800) commit 967865393c08f53c70da9c8cd2824978f2601aa4 tree 3a195e349958669be90c6ec0c80da2f1c5a5dd7e tree parent 46e3ed88ccc738e6b99d0fc957617b4345392237 commit | commitdiff Plug memory leak in doLoadModule() (cherry picked from cf7ca9d09cba14d107152a5179de38e5ef7bd784 commit) http://gitweb.freedesktop.org/?p=xorg/xserver.git;a=commit;h=8a42af6a935b1cf0e15102e986bb527f4fab31a8 Check for clientGone before sending events from XFixes (bug #1753). author Keith Packard <keithp@neko.keithp.com> Mon, 19 Feb 2007 23:28:37 +0000 (15:28 -0800) committer Keith Packard <keithp@neko.keithp.com> Mon, 19 Feb 2007 23:28:37 +0000 (15:28 -0800) commit 8a42af6a935b1cf0e15102e986bb527f4fab31a8 tree 47e7b514da212dbe2f4571f9d2797829fbd3e038 tree parent 4660eaaffb36f526f71d5847ae1309c10ee133c6 commit | commitdiff Check for clientGone before sending events from XFixes (bug #1753). Freeing resources during client closedown can cause cursor changes which attempt to send cursor events through the XFixes extension; a client in the process of closing down has no file to send events to, causing a crash when this path is hit. http://gitweb.freedesktop.org/?p=xorg/xserver.git;a=commit;h=876b806ec09d5ff0c6cd19df91006c4eefedfaa6 Reapply patch to fix AMD CPU detection author Soren Sandmann Pedersen <ssp@dhcp83-218.boston.redhat.com> Tue, 6 Feb 2007 22:16:23 +0000 (17:16 -0500) committer Soren Sandmann Pedersen <ssp@dhcp83-218.boston.redhat.com> Tue, 6 Feb 2007 22:16:23 +0000 (17:16 -0500) commit 876b806ec09d5ff0c6cd19df91006c4eefedfaa6 tree 0c30b71f2adbbea5641bc4611c6360e23a2e172d tree parent 13568d2aa43da4216bbcb46e1125ff28c323ac54 commit | commitdiff Reapply patch to fix AMD CPU detection http://gitweb.freedesktop.org/?p=xorg/xserver.git;a=commit;h=a23b0b069cac8a48e2b306b2095515d75f647705 Typo fix. author Adam Jackson <ajax@benzedrine.nwnk.net> Mon, 12 Feb 2007 22:50:00 +0000 (17:50 -0500) committer Adam Jackson <ajax@benzedrine.nwnk.net> Mon, 12 Feb 2007 22:50:00 +0000 (17:50 -0500) commit a23b0b069cac8a48e2b306b2095515d75f647705 tree 44d98fa9da8b879e4d610a9fc484804a418d217c tree parent d21c95f80bdba2f29eedd57fb0b00e580391c08e commit | commitdiff Typo fix. Reproducible: Always
Created attachment 113728 [details, diff] 1.2.0-check-for-clientgone.patch
Created attachment 113729 [details, diff] 1.2.0-fix-amd-cpu-detection.patch
Created attachment 113731 [details, diff] 1.2.0-plug-memory-leak-in-doloadmodule.patch
Created attachment 113732 [details, diff] 1.2.0-properly-free-device-devprivates-memory-leak-fix.patch
Created attachment 113733 [details, diff] 1.2.0-typo-fix.patch
Created attachment 113734 [details, diff] 1.2.0-zero-out-client-devprivates-on-allocation.patch
Created attachment 113736 [details, diff] xorg-server-1.2.0-r2.patch Patch against xorg-server-1.2.0-r2.ebuild.
Ah, thanks for this. I'll mark it for inclusion, but if the 1.3 server is release soon and contains most of these fixes, they might never get added.
joshuabaergen: Could you please make a -r4 ebuild with these since I haven't seen any move to 1.3 yet? My X was up to 800Mb RSS yesterday (and this box only has 1Gb of memory, so I was deep into swap).
my xorg-server (1.2.0-r2 with ati driver) is using an incredible amount of memory. this is what i get from top: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 5430 root 15 0 1030m 426m 10m R 1.0 42.5 81:42.09 X am i in the right place? is this patch intended to fix the problem i'm seeing?
(In reply to comment #10) > am i in the right place? is this patch intended to fix the problem i'm seeing? > Try updating to xorg-server-1.3 and see if that fixes your issue.
Thanks for these patches. Xorg-server-1.2.0-r2 is devouring a stunning amount of memory for me as well.
i'm now running x11-base/xorg-server-1.3.0.0, amd64 with ati driver still seems to use up a lot of memory. my last X session with 1.3.0 lasted at least a few weeks, and ended with my system hanging (no remote access, no pinging it, nothing). not sure if X was to blame for the hang, but it was using close to 1GB of memory at the time. do you have a good test script for reproducing this? my memory usage seems to jump up the most (and not come back down) when i do a lot of browsing or use picasa under wine -- anything where i'm viewing a lot of images. although i'm not familiar enough with X to know what image-displaying function that might be, a possibility for reproducing this problem more quickly than running X for several weeks would be to randomly generate and show jpg images.
Just FYI, I dug the 1.3.0.0 sources, this includes 2 patches which are filed at comment #1 and #3. Others are not applied ATM, I guess.
I've remerged 1.3.0.0 w/ #2, 4, 5, 6 patches, it just works here. And memory use looks like smaller than vanilla.
can we get the fixes in #4 and #6 included in the portage please? I am hitting this too after days of uptime.
Just committed to 1.3.0.0, thanks for your patience!
Was the release intentionally not bumped?
(In reply to comment #18) > Was the release intentionally not bumped? Yes. Not enough people have been having problems with this (as evidenced by the short CC list on this bug) to make everyone recompile for it. Plus, I want to stabilize 1.3 in the next week or so with these fixes.