Page MenuHomeFreeBSD

KASAN: Disable the sanitizer in stack_capture()
ClosedPublic

Authored by markj on May 5 2021, 2:44 PM.
Tags
None
Referenced Files
Unknown Object (File)
Sat, Apr 19, 11:13 AM
Unknown Object (File)
Sat, Apr 19, 8:33 AM
Unknown Object (File)
Sat, Apr 12, 9:08 AM
Unknown Object (File)
Fri, Apr 11, 6:24 AM
Unknown Object (File)
Tue, Apr 8, 9:18 AM
Unknown Object (File)
Sun, Mar 30, 8:40 AM
Unknown Object (File)
Mar 4 2025, 7:07 PM
Unknown Object (File)
Feb 23 2025, 2:28 PM
Subscribers

Details

Summary

When unwinding the stack we may descend into a poisoned region of the
kernel stack. I don't see a good way to handle this except to disable
the sanitizer entirely.

Define __nosanitizeaddress depending on whether we're in the kernel or
not, since KASAN and ASAN require different attributes. Add
__nosanitizememory while I'm here.

A further complication is that clang raises an error if
no_sanitize("kernel-address") is applied to a global variable rather
than a function:

error: 'no_sanitize' attribute only applies to functions and methods

So, modify linker_set.h to apply the attribute only if we're not in the
kernel.

Diff Detail

Repository
rS FreeBSD src repository - subversion
Lint
Lint Passed
Unit
No Test Coverage
Build Status
Buildable 39028
Build 35917: arc lint + arc unit

Event Timeline

markj requested review of this revision.May 5 2021, 2:44 PM
sys/x86/x86/stack_machdep.c
73

I have no idea what is going on there, but don't all called functions require the same handling, e.g. stack_zero? Is it happen to work because struct stack itself is 'not poisoned'?

sys/x86/x86/stack_machdep.c
73

stack_zero() is fine, it's the inline dereference of the frame pointer that potentially causes problems.

Mostly on principle "I do not see anything wrong"

This revision is now accepted and ready to land.May 5 2021, 8:02 PM