segfault in configuring linux-restricted modules

[MaxNegro]

I upgraded to edgy a couple of days ago, and on one of my machines, apt-get is unable to configure linux-restricted-modules due to a segfault in executing depmod -a -F "System.map-2.6.17-10-generic" "2.6.17-generic".

Can't say why it's acting like that, anyway I recompiled module-init-tools with debugging symbols and obtained the following:

(gdb) handle SIG33 pass nostop noprint
Signal Stop Print Pass to program Description
SIG33 No No Yes Real-time event 33
(gdb) run -a -q -F /boot/System.map-"2.6.17-10-generic" "2.6.17-10-generic"
Starting program: /sbin/depmod -a -q -F /boot/System.map-"2.6.17-10-generic" "2.6.17-10-generic"

Program received signal SIGSEGV, Segmentation fault.
0x0804a1e4 in main (argc=6, argv=0xbfca9fd4) at ../depmod.c:151
151 ptr = strchr(line, '\n');
(gdb) thread apply all bt
(gdb)
(gdb) backtrace
#0 0x0804a1e4 in main (argc=6, argv=0xbfca9fd4) at ../depmod.c:151

Hope it helps.

[MaxNegro (maxnegro)]

Ok, found the cause. I misconfigured lilo.conf and so System.map got corrupted. Anyway I think assigning a char to a pointer that could be null is still a bug. (maybe using strchrnul could be the case?) :)

[Alan Jenkins (aj504)]

When module-init-tools was checked into the git repo, two months after you hit this bug, it looked like this

175 ptr = strchr(line, '\n');
176 *ptr = '\0';

So I guess gdb got the line number off by one.

But the current code checks ptr is non-NULL before assigning to it, so this shouldn't happen any more :-).