Re: [PATCH] Revert Intel CET changes to __jmp_buf_tag (Bug 22743)

[Carlos O'Donell <carlos@redhat.com>]

On 01/24/2018 09:33 PM, H.J. Lu wrote:
> On Wed, Jan 24, 2018 at 8:53 PM, Carlos O'Donell <carlos@redhat.com> wrote:
>> On 01/24/2018 05:48 PM, Dmitry V. Levin wrote:
>>> I'm afraid by Monday it will be too late for 2.27 as we will get very
>>> little testing before the release.
>>  Before reverting:
>>
>> [carlos@athas tst-cleanup1]$ /home/carlos/build/glibc/elf/ld.so --library-path /home/carlos/build/glibc:/home/carlos/build/glibc/elf:/home/carlos/build/glibc/dlfcn:/home/carlos/build/glibc/nptl ./tst-cleanup1
>> ch (3)
>> ch (2)
>> ch (1)
>> Didn't expect signal from child: got `Segmentation fault'
>>
>> After reverting:
>>
>> [carlos@athas tst-cleanup1]$ /home/carlos/build/glibc-reverted/elf/ld.so --library-path /home/carlos/build/glibc-reverted:/home/carlos/build/glibc-reverted/elf:/home/carlos/build/glibc-reverted/dlfcn:/home/carlos/build/glibc-reverted/nptl ./tst-cleanup1
>> ch (3)
>> ch (2)
>> ch (1)
>>
>> ~~~ Commit message ~~~
>> In commit cba595c350e52194e10c0006732e1991e3d0803b and commit
>> f81ddabffd76ac9dd600b02adbf3e1dac4bb10ec, ABI compatibility with
>> applications was broken by increasing the size of the on-stack
>> allocated __pthread_unwind_buf_t beyond the oringal size.
>> Applications only have the origianl space available for
>> __pthread_unwind_register, and __pthread_unwind_next to use,
>> any increase in the size of __pthread_unwind_buf_t causes these
>> functions to write beyond the original structure into other
>> on-stack variables leading to segmentation faults in common
>> applications like vlc. The only workaround is to version those
>> functions which operate on the old sized objects, but this must
>> happen in glibc 2.28.
>>
>> Thank you to Andrew Senkevich, H.J. Lu, and Aurelien Jarno, for
>> submitting reports and tracking the issue down.
>>
>> The commit reverts the above mentioned commits and testing on
>> x86_64 shows that the ABI compatibility is restored. A tst-cleanup1
>> regression test linked with an older glibc now passes when run
>> with the newly built glibc. Previously a tst-cleanup1 linked with
>> an older glibc would segfault when run with an affected glibc build.
>>
>> Tested on x86_64 with no regressions.
>>
>> Signed-off-by: Carlos O'Donell <carlos@redhat.com>
>> ~~~
>>
>> Patch attached.
>>
>> OK to commit?
>>
>> This fixes the last blocker for glibc 2.27.
> 
> Please don't revert my patch.  Please try this patch:
> 
> https://sourceware.org/git/?p=glibc.git;a=commit;h=4b7fc470a6740808b41502d7431f91805e272d26
> 
> instead.  I will clean it up and submit it tomorrow.

This is unacceptable. It adds a new symbol version and we froze the
ABI at the start of the month. You cannot work these fixes into 2.27
with a new symbol version, it must wait for 2.28.

-- 
Cheers,
Carlos.