On Sat, Nov 10, 2018 at 8:15 AM, Jordan Glover
<Golden_Miller83@xxxxxxxxxxxxx> wrote:
> Hello,
>
> Commit 6c3b7af1c975b87b86dcb2af233d1ae21eb05107 ("kyber: add tracepoints")[1] causes write beyond size of object. This was detected by "FORTIFY_SOURCE intra-object overflow checking"[2] feature which is part of linux-hardened out-of-tree patchset designed to catch such errors.
>
> The specific error is:
>
> In file included from ./include/linux/bitmap.h:9,
> from ./include/linux/cpumask.h:12,
> from ./arch/x86/include/asm/cpumask.h:5,
> from ./arch/x86/include/asm/msr.h:11,
> from ./arch/x86/include/asm/processor.h:21,
> from ./arch/x86/include/asm/cpufeature.h:8,
> from ./arch/x86/include/asm/thread_info.h:53,
> from ./include/linux/thread_info.h:38,
> from ./arch/x86/include/asm/preempt.h:7,
> from ./include/linux/preempt.h:81,
> from ./include/linux/rcupdate.h:40,
> from ./include/linux/rculist.h:11,
> from ./include/linux/pid.h:5,
> from ./include/linux/sched.h:14,
> from ./include/linux/blkdev.h:5,
> from block/kyber-iosched.c:21:
> In function ‘strlcpy’,
> inlined from ‘perf_trace_kyber_latency’ at ./include/trace/events/kyber.h:14:1:
> ./include/linux/string.h:310:4: error: call to ‘__write_overflow’ declared with attribute error: detected write beyond size of object passed as 1st parameter
> __write_overflow();
> ^~~~~~~~~~~~~~~~~~
> In function ‘strlcpy’,
> inlined from ‘trace_event_raw_event_kyber_latency’ at ./include/trace/events/kyber.h:14:1:
> ./include/linux/string.h:310:4: error: call to ‘__write_overflow’ declared with attribute error: detected write beyond size of object passed as 1st parameter
> __write_overflow();
> ^~~~~~~~~~~~~~~~~~
> make[1]: *** [scripts/Makefile.build:293: block/kyber-iosched.o] Error 1
> make: *** [Makefile:1063: block] Error 2
> make: *** Waiting for unfinished jobs....
>
> Using 'strlcpy' function is generally not recommended[3][4].
Due to the macros, this was a little tricky to find, but it looks like
a cut/paste typo:
#define DOMAIN_LEN 16
#define LATENCY_TYPE_LEN 8
strlcpy(__entry->domain, domain, DOMAIN_LEN);
strlcpy(__entry->type, type, DOMAIN_LEN);
This should use strscpy() regardless, and should use sizeof(dst)
instead of separate literals. The primary bug is using DOMAIN_LEN for
__entry->type when it is actually LATENCY_TYPE_LEN bytes.
Can you build a patch for this? I'm happy to review.
Thanks for finding this!
-Kees
>
> Jordan
>
> [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=v4.20-rc1&id=6c3b7af1c975b87b86dcb2af233d1ae21eb05107
>
> [2] https://github.com/anthraxx/linux-hardened/commit/9460692de8eb53fd62d59f564eba215e7c03a34b
>
> [3] https://lwn.net/Articles/763641/
>
> [4] https://outflux.net/slides/2018/lss/danger.pdf
--
Kees Cook
