grsecurity forums

by **KDE** » Sat Jul 21, 2012 4:46 am

Chrome 21.0.1180.49 often crashes at start-up on kernel 3.4.5-hardened

PAX: kernel memory leak attempt detected from (nil) (<null>) (4096 bytes)
Call Trace:
[<ffffffff810ea0e1>] ? 0xffffffff810ea0e1
[<ffffffff814fa4f5>] ? 0xffffffff814fa4f5
[<ffffffff810dd0e4>] ? 0xffffffff810dd0e4
[<ffffffff810aff4c>] ? 0xffffffff810aff4c
[<ffffffff810b0ed9>] ? 0xffffffff810b0ed9
[<ffffffff810afe50>] ? 0xffffffff810afe50
[<ffffffff8102e929>] ? 0xffffffff8102e929
[<ffffffff810e1445>] ? 0xffffffff810e1445
[<ffffffff814f7835>] ? 0xffffffff814f7835
[<ffffffff814f75fb>] ? 0xffffffff814f75fb
[<ffffffff810e1eeb>] ? 0xffffffff810e1eeb
[<ffffffff810e205e>] ? 0xffffffff810e205e
[<ffffffff814f7f32>] ? 0xffffffff814f7f32
[<ffffffff810e174f>] ? 0xffffffff810e174f
[<ffffffff814f7f58>] ? 0xffffffff814f7f58

by **PaX Team** » Sat Jul 21, 2012 5:56 am

can you send me the *exact* vmlinux image that produced this report? or even better, enable KALLSYMS (grsec disables it on purpose) and provide us with a backtrace using symbols please? also try something newer like 3.4.6

.

by **KDE** » Sat Jul 21, 2012 6:05 am

I will try with 3.4.6 when it will be available on Gentoo.

by **PaX Team** » Sun Jul 22, 2012 6:57 pm

thanks for the decoded oops, it seems that we'll have to dig deeper, so can you strace -f chrome and send the logs to me please? (perhap sanitize them, i really only need to see the syscalls that are related to the one triggering the failing aio read)

by **KDE** » Mon Jul 23, 2012 5:26 am

I can't reproduce crash with strace.

by **PaX Team** » Mon Jul 23, 2012 7:30 pm

KDE wrote:I can't reproduce crash with strace.

hmm, that doesn't make life easier

. can you send me your kernel config? also what do you need to trigger this? just attempt to start chrome? also, is it the google build or something you compiled?

by **PaX Team** » Thu Aug 02, 2012 9:26 pm

so i've been trying to reproduce this with different kernel and chrome versions to no avail so far. looking at the source also makes this an 'impossible' bug (kmap cannot return NULL) so i'm thinking there's something weird going on with your kernel builds. can you send me a vmlinux (not bzImage) and corresponding backtrace please?

by **KDE** » Fri Aug 03, 2012 2:23 am

It seems it is already fixed in 3.4.7

grsecurity forums

Chrome_HistoryT: memory leak attempt

Chrome_HistoryT: memory leak attempt

Re: Chrome_HistoryT: memory leak attempt

Re: Chrome_HistoryT: memory leak attempt

Re: Chrome_HistoryT: memory leak attempt

Re: Chrome_HistoryT: memory leak attempt

Re: Chrome_HistoryT: memory leak attempt

Re: Chrome_HistoryT: memory leak attempt

Re: Chrome_HistoryT: memory leak attempt