| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
I don't think I'm an appropriate reviewer for this CL.
hongchan@, would you review this please?
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Looks like the suggested changes is closely related to the Blink allocator.
Perhaps nuskos@ is the right expert?
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Code-Review | -1 |
I think more investigation needs to be done here.
(BRP) is enabled in the BufferPartition, it can shift the object's address
by adding metadata at the start of the memory slot, violating the alignmentThis isn't true, metadata is at the end of the slot, it shouldn't move the address at all.
which has BRP disabled to support strict alignment. If the ArrayBufferPartitionThis is disabling a pretty big protection, I wouldn't like to do this.
BufferPartition. BufferAlignedFree is also updated to use FreeInUnknownRoot
so it can handle pointers from either partition.This is a pretty big potential performance regression. You'd need to run speedometer3 and jetstream2 and ideally some sort of audio heavy benchmark if one exists?
**This fix was generated by the Crash Autofix pipeline for the crash http://crash/3e05f4efd8e421fc. Please see go/crash-autofix for more details.**This doesn't look like a correct fix at all.
1) It made up the fact about metadata
2) It tries to fix it by disabling a pretty large UaF security protection.
3) It adds performance overhead for all users of BufferAlignedFree (although there likely aren't many).
This should likely go in the pile of bad outcomes to train the AI to improve.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |