Write stacks correctly during stack overflows #78
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
This is a WIP, I need to test it in real world scenarios and add tests too. |
gabrielesvelto
force-pushed
the
stack-overflow
branch
3 times, most recently
from
a604403 to
d8179ce
Compare
|
This is better, it works in Firefox so I will be able to add tests soon. There's still a couple of things that need to be hashed out:
|
gabrielesvelto
force-pushed
the
stack-overflow
branch
5 times, most recently
from
f92b998 to
cc90c1e
Compare
gabrielesvelto
force-pushed
the
stack-overflow
branch
2 times, most recently
from
a2c68a5 to
d1c35f4
Compare
When encountering a stack overflow we often crash accessing the guard page. The logic assumed that wherever the stack pointer was so was the stack, but this lead the writer to dump the guard page in these cases. This patch changes the logic to inspect the properties of the mapping that appears to correspond to the stack and - if it looks like a guard page - look for the actual stack instead. This change also removes the double limitation we had when retrieving stacks on Linux: previously the logic would only grab the first 32 KiB of each stack before checking for user-specified limits. Now only the user-specified limits are enforced and - if not present - the full stack is stored in the minidump. This brings the behavior in line with minidumps generated on Windows by windbg.dll. This fixes rust-minidump#24
gabrielesvelto
force-pushed
the
stack-overflow
branch
2 times, most recently
from
d14dfe9 to
04bfdb7
Compare
|
This is as good as it gets, I tried to match the original intent, removing the redundant limit on the amount of stack we captured while ensuring we always grab a stack, provided the overflow is within 1MiB of where it should be as that's the size of guard pages in Linux ATM. |
|
@Jake-Shadle would you have time to take a look? |
|
Oh sorry, this completely fell off my radar, will review tomorrow! |
|
Thanks! |
Jake-Shadle
approved these changes
May 26, 2023
Comment on lines
+340
to
352
|
|
||
| pub fn is_executable(&self) -> bool { | ||
| self.permissions.contains(MMPermissions::EXECUTE) | ||
| } | ||
|
|
||
| pub fn is_readable(&self) -> bool { | ||
| self.permissions.contains(MMPermissions::READ) | ||
| } | ||
|
|
||
| pub fn is_writable(&self) -> bool { | ||
| self.permissions.contains(MMPermissions::WRITE) | ||
| } | ||
| } |
There was a problem hiding this comment.
These could be inline, but that's just a nit.
No rush at all. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When encountering a stack overflow we often crash accessing the guard page. The logic assumed that wherever the stack pointer was so was the stack, but this lead the writer to dump the guard page in these cases. This patch changes the logic to inspect the properties of the mapping that appears to correspond to the stack and - if it looks like a guard page - look for the actual stack instead.
This fixes #24