On 4/10/18 12:17 PM, Cyrill Gorcunov wrote:
On Tue, Apr 10, 2018 at 11:28:13AM -0700, Yang Shi wrote:
At the first glance, it looks feasible to me. Will look into deeper
A further look told me this might be *not* feasible.

It looks the new lock will not break check_data_rlimit since in my patch
both start_brk and brk is protected by mmap_sem. The code flow might look
like below:

CPU A                             CPU B
--------                       --------
prctl                               sys_brk
check_data_rlimit           check_data_rlimit (need mm->start_brk)
                                       set brk
down_write                    up_write
set start_brk
set brk

If CPU A gets the mmap_sem first, it will set start_brk and brk, then CPU B
will check with the new start_brk. And, prctl doesn't care if sys_brk is run
before it since it gets the new start_brk and brk from parameter.

If we protect start_brk and brk with the new lock, sys_brk might get old
start_brk, then sys_brk might break rlimit check silently, is that right?

So, it looks using new lock in prctl and keeping mmap_sem in brk path has
race condition.
I fear so. The check_data_rlimit implies that all elements involved into
validation (brk, start_brk, start_data, end_data) are not changed unpredicably
until written back into mm. In turn if we guard start_brk,brk only (as
it is done in the patch) the check_data_rlimit may pass on wrong data
I think. And as you mentioned the race above exact the example of such
situation. I think for prctl case we can simply left use of mmap_sem
as it were before the patch, after all this syscall is really in cold
path all the time.

The race condition is just valid when protecting start_brk, brk, start_data and end_data with the new lock, but keep using mmap_sem in brk path.

So, we should just need make a little tweak to have mmap_sem protect start_brk, brk, start_data and end_data, then have the new lock protect others so that we still can remove mmap_sem in proc as the patch is aimed to do.



Reply via email to