tags 302378 patch pending thanks And valgrind reveals all in the end. Attached is a patch which fixes up a rounding error in the use of malloc_array(); this is probably a bit wasteful compared to the original code, but at least it never allocates too little and gives us malloc corruption.
I'll get this over to the security team asap; in the meantime, the patch is here for people who can't wait for the official builds. Cheers, -- Steve Langasek postmodern programmer On Fri, Apr 08, 2005 at 09:48:53AM +0200, Rasmus Bøg Hansen wrote: > Andrew Bartlett <[EMAIL PROTECTED]> hit the keyboard. > Afterwards the following was on the screen: > > > On Fri, 2005-04-08 at 01:08 +0200, Rasmus Bøg Hansen wrote: > >> Hi again > >> > >> Valgrind is not in woody, so I installed it from backports, and made > >> the following trace: > >> > >> valgrind -v -- smbd -i > /tmp/valgrind.log 2>&1 > >> > >> Output is here: > >> > >> http://www.amagerkollegiet.dk/~moffe/valgrind.log > > > > Sorry, that doesn't help. We will probably need --num-callers=32, and > > we need to show it crashing, not just running and exiting once (which is > > what -i does). > > > > Internally, Samba has been modified to assist in this, so if you run > > with the valgrind --trace-children option, but not smbd's -i, the > > valgrind errors will end up in the normal Samba logfiles (we redirect > > stderr). > > Ok, I now did: > > valgrind -v --num-callers=32 --trace-children=yes --log-file=valgrind.log -- > smbd > > I do not really understand (perhaps due to my limited knowledge of > valgrind) why smbd crashes when run normally; when run through > valgrind it seems to run without crashing. > > Nonetheless, the valgrind logfile is here: > > http://www.amagerkollegiet.dk/~moffe/valgrind.log.pid2456
signature.asc
Description: Digital signature
