[GIT PULL] blob-stacking updates for security-next

Casey Schaufler casey at schaufler-ca.com
Fri Jan 11 17:29:36 UTC 2019


On 1/11/2019 2:38 AM, Tetsuo Handa wrote:
> On 2019/01/09 6:35, Kees Cook wrote:
>> Hi James,
>>
>> Please pull these blob-stacking changes for security-next.
>>
>> Thanks!
>>
>> -Kees
>>
>> The following changes since commit bfeffd155283772bbe78c6a05dec7c0128ee500c:
>>
>>   Linux 5.0-rc1 (2019-01-06 17:08:20 -0800)
>>
>> are available in the Git repository at:
>>
>>   https://git.kernel.org/pub/scm/linux/kernel/git/kees/linux.git tags/blob-stacking-security-next
>>
>> for you to fetch changes up to a5e2fe7ede1268d2f80fe49ca1f717d0e3750995:
>>
>>   TOMOYO: Update LSM flags to no longer be exclusive (2019-01-08 13:18:45 -0800)
>>
> And syzbot already found a bug.
> This is occurring immediately after memory allocation failure for cred object.
> We need to be prepared for free() function being called when alloc() function failed.
>
> [   59.992498][ T8010] FAULT_INJECTION: forcing a failure.
> [   59.992498][ T8010] name failslab, interval 1, probability 0, space 0, times 1
> [   60.005214][ T8010] CPU: 0 PID: 8010 Comm: syz-executor178 Not tainted 5.0.0-rc1-next-20190111 #10
> [   60.014337][ T8010] Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
> [   60.024383][ T8010] Call Trace:
> [   60.027657][ T8010]  dump_stack+0x1db/0x2d0
> [   60.063731][ T8010]  should_fail.cold+0xa/0x14
> [   60.089894][ T8010]  __should_failslab+0x121/0x190
> [   60.094810][ T8010]  should_failslab+0x9/0x14
> [   60.099411][ T8010]  __kmalloc+0x2dc/0x740
> [   60.124293][ T8010]  security_prepare_creds+0x123/0x190
> [   60.129644][ T8010]  prepare_creds+0x3c4/0x510
> [   60.149852][ T8010]  __x64_sys_capset+0x58c/0x9b0
> [   60.185347][ T8010]  do_syscall_64+0x1a3/0x800
> [   60.206747][ T8010]  entry_SYSCALL_64_after_hwframe+0x49/0xbe

I'm staring at the code and trying to see what this is telling me.
There is a failure check immediately after the allocation. If any
of the modules fail in their hooks, the whole thing is freed.



More information about the Linux-security-module-archive mailing list