[perf-prof] Fix erroneous code offsets in unwinding info

[perf-prof] Fix erroneous code offsets in unwinding info

The unwinding information we emit wrongly encodes code locations as relative
offsets. If we look at the .eh_frame section of shared object generated by "perf
inject" using "objdump -g":

~~~
00000000 0000000000000018 00000000 CIE
(snip)
0000001c 0000000000000028 00000020 FDE cie=00000000 pc=fffffffffffffee8..00000000000017f8
(snip)
00000048 ZERO terminator
~~~

We can see the range that the FDE entry covers is incorrect, it should point to
where the .text section is, at address 0x40 on a 64-bit architecture.

The reason for this was that the PerfJitLogger logs a code size that is
different from the one we've used when encoding the unwinding information. The
logger will ignore the safepoint table while the unwinding info assumes it is
part of the code.

BUG=

Review-Url: https://codereview.chromium.org/2790403002
Cr-Commit-Position: refs/heads/master@{#44378}
Committed: https://chromium.googlesource.com/v8/v8/+/21f064fcdceb162524ab1bc5ed2b8ff6a803c7b4

[Pierre Langlois, 2017-04-04 13:08:01]

Description was changed from

==========
[perf-prof] Fix erroneous code offsets in unwinding info

The unwinding information we emit wrongly encodes code locations as relative
offsets. If we look at the .eh_frame section of shared object generated by "perf
inject" using "objdump -g":

~~~
00000000 0000000000000018 00000000 CIE
(snip)
0000001c 0000000000000028 00000020 FDE cie=00000000
pc=fffffffffffffee8..00000000000017f8
(snip)
00000048 ZERO terminator
~~~

We can see the range that the FDE entry covers is incorrect, it should point to
where the .text section is, at address 0x40 on a 64-bit architecture.

The reason for this was that the PerfJitLogger logs a code size that is
different from the one we've used when encoding the unwinding information. The
logger will ignore the safepoint table while the unwinding info assumes it is
part of the code.

BUG=
==========

to

==========
[perf-prof] Fix erroneous code offsets in unwinding info

The unwinding information we emit wrongly encodes code locations as relative
offsets. If we look at the .eh_frame section of shared object generated by "perf
inject" using "objdump -g":

~~~
00000000 0000000000000018 00000000 CIE
(snip)
0000001c 0000000000000028 00000020 FDE cie=00000000
pc=fffffffffffffee8..00000000000017f8
(snip)
00000048 ZERO terminator
~~~

We can see the range that the FDE entry covers is incorrect, it should point to
where the .text section is, at address 0x40 on a 64-bit architecture.

The reason for this was that the PerfJitLogger logs a code size that is
different from the one we've used when encoding the unwinding information. The
logger will ignore the safepoint table while the unwinding info assumes it is
part of the code.

BUG=
==========

[Pierre Langlois, 2017-04-04 13:08:01]

pierre.langlois@arm.com changed reviewers:
+ danno@chromium.org, jarin@chromium.org

[danno, 2017-04-04 13:14:05]

Thanks for the patch! Nice catch. Kudos for digging into the perf support and
fixing some of the glitches. LGTM.

[Pierre Langlois, 2017-04-04 13:21:45]

On 2017/04/04 13:14:05, danno wrote:
> Thanks for the patch! Nice catch. Kudos for digging into the perf support and
> fixing some of the glitches. LGTM.

Thanks for the quick review! The perf JIT support is very useful.

[Pierre Langlois, 2017-04-04 13:21:50]

The CQ bit was checked by pierre.langlois@arm.com

[commit-bot: I haz the power, 2017-04-04 13:22:03]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2017-04-04 14:24:47]

CQ is committing da patch.

Bot data: {"patchset_id": 1, "attempt_start_ts": 1491312110276770, "parent_rev":
"c766727ae69d43db66415a3483acc0e846b444d7", "commit_rev":
"21f064fcdceb162524ab1bc5ed2b8ff6a803c7b4"}

[commit-bot: I haz the power, 2017-04-04 14:24:54]

Description was changed from

==========
[perf-prof] Fix erroneous code offsets in unwinding info

The unwinding information we emit wrongly encodes code locations as relative
offsets. If we look at the .eh_frame section of shared object generated by "perf
inject" using "objdump -g":

~~~
00000000 0000000000000018 00000000 CIE
(snip)
0000001c 0000000000000028 00000020 FDE cie=00000000
pc=fffffffffffffee8..00000000000017f8
(snip)
00000048 ZERO terminator
~~~

We can see the range that the FDE entry covers is incorrect, it should point to
where the .text section is, at address 0x40 on a 64-bit architecture.

The reason for this was that the PerfJitLogger logs a code size that is
different from the one we've used when encoding the unwinding information. The
logger will ignore the safepoint table while the unwinding info assumes it is
part of the code.

BUG=
==========

to

==========
[perf-prof] Fix erroneous code offsets in unwinding info

The unwinding information we emit wrongly encodes code locations as relative
offsets. If we look at the .eh_frame section of shared object generated by "perf
inject" using "objdump -g":

~~~
00000000 0000000000000018 00000000 CIE
(snip)
0000001c 0000000000000028 00000020 FDE cie=00000000
pc=fffffffffffffee8..00000000000017f8
(snip)
00000048 ZERO terminator
~~~

We can see the range that the FDE entry covers is incorrect, it should point to
where the .text section is, at address 0x40 on a 64-bit architecture.

The reason for this was that the PerfJitLogger logs a code size that is
different from the one we've used when encoding the unwinding information. The
logger will ignore the safepoint table while the unwinding info assumes it is
part of the code.

BUG=

Review-Url: https://codereview.chromium.org/2790403002
Cr-Commit-Position: refs/heads/master@{#44378}
Committed:
https://chromium.googlesource.com/v8/v8/+/21f064fcdceb162524ab1bc5ed2b8ff6a80...
==========

[commit-bot: I haz the power, 2017-04-04 14:24:55]

Committed patchset #1 (id:1) as
https://chromium.googlesource.com/v8/v8/+/21f064fcdceb162524ab1bc5ed2b8ff6a80...