[PING] [PATCH 0/3] Fix for PR 17808 and some related changes

[PING] [PATCH 0/3] Fix for PR 17808 and some related changes

[Andreas Arnez]

Ping:

  https://sourceware.org/ml/gdb-patches/2015-01/msg00424.html

Pedro commented already, as shown below.  Mark has not replied so far.

Note that the PR is on the TODO list for 7.9.  In my opinion at least
patch 1 should go in 7.9; it's fairly harmless and should fix the PR.
Patch 2 fixes a similar problem, but there might be a test gap, because
I currently don't have access to appropriate test hardware (with AVX and
AVX-512).  And with patch 3 there's a slight chance of introducing bogus
warnings when reading core files.  Thus I'd suggest to push all three
patches upstream, but only patch 1 into 7.9.  OK?

On Fri, Jan 23 2015, Pedro Alves wrote:

> On 01/15/2015 03:21 PM, Andreas Arnez wrote:
>> As seen in PR 17808, a test case with a forged (invalid) core file can
>> crash GDB with an assertion failure when a register section has the
>> wrong size.  This patch series is about improving GDB's behavior in
>> such cases: Patch 1 fixes the reported problem for i386 as well as for
>> other targets; patch 2 avoids out-of-bounds accesses when an x86
>> ".reg-xstate" section is too short, and patch 3 adds a warning when a
>> core file register section is larger than expected.
>> 
>> This is based on the "lazy approach" suggested here:
>> 
>>   https://sourceware.org/ml/gdb-patches/2015-01/msg00229.html
>> 
>> OK to apply?
>
> This looks good to me.  Mark, any comments?
>
> Thanks,
> Pedro Alves

Re: [PING] [PATCH 0/3] Fix for PR 17808 and some related changes

[Joel Brobecker]

On Thu, Jan 29, 2015 at 04:58:07PM +0100, Andreas Arnez wrote:
> Ping:
> 
>   https://sourceware.org/ml/gdb-patches/2015-01/msg00424.html
> 
> Pedro commented already, as shown below.  Mark has not replied so far.
> 
> Note that the PR is on the TODO list for 7.9.  In my opinion at least
> patch 1 should go in 7.9; it's fairly harmless and should fix the PR.
> Patch 2 fixes a similar problem, but there might be a test gap, because
> I currently don't have access to appropriate test hardware (with AVX and
> AVX-512).  And with patch 3 there's a slight chance of introducing bogus
> warnings when reading core files.  Thus I'd suggest to push all three
> patches upstream, but only patch 1 into 7.9.  OK?

Let's give it another week for additional comments (or request for
more time to review), and then push it to master.

For 7.9, let's confirm your assessment with Pedro. It sounds pretty
good to me.

-- 
Joel

Re: [PING] [PATCH 0/3] Fix for PR 17808 and some related changes

[Pedro Alves]

On 02/02/2015 04:50 AM, Joel Brobecker wrote:
> On Thu, Jan 29, 2015 at 04:58:07PM +0100, Andreas Arnez wrote:
>> Ping:
>>
>>   https://sourceware.org/ml/gdb-patches/2015-01/msg00424.html
>>
>> Pedro commented already, as shown below.  Mark has not replied so far.
>>
>> Note that the PR is on the TODO list for 7.9.  In my opinion at least
>> patch 1 should go in 7.9; it's fairly harmless and should fix the PR.
>> Patch 2 fixes a similar problem, but there might be a test gap, because
>> I currently don't have access to appropriate test hardware (with AVX and
>> AVX-512).  And with patch 3 there's a slight chance of introducing bogus
>> warnings when reading core files.  Thus I'd suggest to push all three
>> patches upstream, but only patch 1 into 7.9.  OK?
> 
> Let's give it another week for additional comments (or request for
> more time to review), and then push it to master.
> 
> For 7.9, let's confirm your assessment with Pedro. It sounds pretty
> good to me.
> 

Sounds good to me, but I have a counter proposal.

How about we push it all to master now?  That'd give it exposure on
both auto testers and on others' machines immediately.  We can always
address any additional comments as follow ups, of course.  The main
difference is that the series would be exposed to testing one
extra week.  Then if we see no fall out, we'd have a little more
confidence pushing to 7.9.

Thanks,
Pedro Alves

Re: [PING] [PATCH 0/3] Fix for PR 17808 and some related changes

[Joel Brobecker]

> Sounds good to me, but I have a counter proposal.
> 
> How about we push it all to master now?  That'd give it exposure on
> both auto testers and on others' machines immediately.  We can always
> address any additional comments as follow ups, of course.  The main
> difference is that the series would be exposed to testing one
> extra week.  Then if we see no fall out, we'd have a little more
> confidence pushing to 7.9.

Sounds even better to me.

Andreas, please push your changes now.

Thank you both!
-- 
Joel

Re: [PING] [PATCH 0/3] Fix for PR 17808 and some related changes

[Andreas Arnez]

On Wed, Feb 04 2015, Joel Brobecker wrote:

>> Sounds good to me, but I have a counter proposal.
>> 
>> How about we push it all to master now?  That'd give it exposure on
>> both auto testers and on others' machines immediately.  We can always
>> address any additional comments as follow ups, of course.  The main
>> difference is that the series would be exposed to testing one
>> extra week.  Then if we see no fall out, we'd have a little more
>> confidence pushing to 7.9.
>
> Sounds even better to me.
>
> Andreas, please push your changes now.

Done.  Thanks for your support!

BTW, do we also want to get the test case upstream that triggered this
PR?  Jan posted it here:

  https://sourceware.org/ml/gdb-patches/2015-01/msg00199.html

Re: [PING] [PATCH 0/3] Fix for PR 17808 and some related changes

[Joel Brobecker]

> >> Sounds good to me, but I have a counter proposal.
> >> 
> >> How about we push it all to master now?  That'd give it exposure on
> >> both auto testers and on others' machines immediately.  We can always
> >> address any additional comments as follow ups, of course.  The main
> >> difference is that the series would be exposed to testing one
> >> extra week.  Then if we see no fall out, we'd have a little more
> >> confidence pushing to 7.9.
> >
> > Sounds even better to me.
> >
> > Andreas, please push your changes now.
> 
> Done.  Thanks for your support!

And, for the record, I have just pushed the patch to gdb-7.9-branch
as well. Given that I was not having my best day yesterday, I re-
tested the patch on x86_64-linux before pushing.

-- 
Joel

[PATCH 0/3] Fix for PR 17808 and some related changes

[Andreas Arnez]

As seen in PR 17808, a test case with a forged (invalid) core file can
crash GDB with an assertion failure when a register section has the
wrong size.  This patch series is about improving GDB's behavior in
such cases: Patch 1 fixes the reported problem for i386 as well as for
other targets; patch 2 avoids out-of-bounds accesses when an x86
".reg-xstate" section is too short, and patch 3 adds a warning when a
core file register section is larger than expected.

This is based on the "lazy approach" suggested here:

  https://sourceware.org/ml/gdb-patches/2015-01/msg00229.html

OK to apply?


Andreas Arnez (3):
  [PR corefiles/17808] Fix internal error when core file section is too
    big
  x86: Use correct .reg-xstate section size
  Warn if core file register section is larger than expected

 gdb/alphanbsd-tdep.c     |  4 +++-
 gdb/amd64-linux-tdep.c   |  2 +-
 gdb/amd64-tdep.c         |  4 ++--
 gdb/armbsd-tdep.c        |  4 +++-
 gdb/corelow.c            |  5 +++++
 gdb/gdbarch.h            |  6 ++++++
 gdb/gdbarch.sh           |  6 ++++++
 gdb/hppa-hpux-tdep.c     |  4 +++-
 gdb/hppaobsd-tdep.c      |  4 +++-
 gdb/i386-linux-tdep.c    |  3 +--
 gdb/i386-tdep.c          |  8 ++++----
 gdb/m68kbsd-tdep.c       |  4 +++-
 gdb/mips-linux-tdep.c    | 16 ++++++++--------
 gdb/mipsnbsd-tdep.c      |  4 +++-
 gdb/mn10300-linux-tdep.c |  8 ++++----
 gdb/regset.h             |  7 +++++++
 16 files changed, 62 insertions(+), 27 deletions(-)

-- 
1.8.4.2

[PING] [PATCH 0/3] Fix for PR 17808 and some related changes

[Andreas Arnez]

Ping:

  https://sourceware.org/ml/gdb-patches/2015-01/msg00424.html

Also, here is Jan's test case that showed the problem:

  https://sourceware.org/ml/gdb-patches/2015-01/msg00199.html

On Thu, Jan 15 2015, Andreas Arnez wrote:

> As seen in PR 17808, a test case with a forged (invalid) core file can
> crash GDB with an assertion failure when a register section has the
> wrong size.  This patch series is about improving GDB's behavior in
> such cases: Patch 1 fixes the reported problem for i386 as well as for
> other targets; patch 2 avoids out-of-bounds accesses when an x86
> ".reg-xstate" section is too short, and patch 3 adds a warning when a
> core file register section is larger than expected.
>
> This is based on the "lazy approach" suggested here:
>
>   https://sourceware.org/ml/gdb-patches/2015-01/msg00229.html
>
> OK to apply?
>
>
> Andreas Arnez (3):
>   [PR corefiles/17808] Fix internal error when core file section is too
>     big
>   x86: Use correct .reg-xstate section size
>   Warn if core file register section is larger than expected
>
>  gdb/alphanbsd-tdep.c     |  4 +++-
>  gdb/amd64-linux-tdep.c   |  2 +-
>  gdb/amd64-tdep.c         |  4 ++--
>  gdb/armbsd-tdep.c        |  4 +++-
>  gdb/corelow.c            |  5 +++++
>  gdb/gdbarch.h            |  6 ++++++
>  gdb/gdbarch.sh           |  6 ++++++
>  gdb/hppa-hpux-tdep.c     |  4 +++-
>  gdb/hppaobsd-tdep.c      |  4 +++-
>  gdb/i386-linux-tdep.c    |  3 +--
>  gdb/i386-tdep.c          |  8 ++++----
>  gdb/m68kbsd-tdep.c       |  4 +++-
>  gdb/mips-linux-tdep.c    | 16 ++++++++--------
>  gdb/mipsnbsd-tdep.c      |  4 +++-
>  gdb/mn10300-linux-tdep.c |  8 ++++----
>  gdb/regset.h             |  7 +++++++
>  16 files changed, 62 insertions(+), 27 deletions(-)