Page MenuHomeFreeBSD

kmem: Check for overflows when rounding up allocation sizes
Needs ReviewPublic

Authored by markj on Jul 24 2024, 9:31 PM.
Tags
None
Referenced Files
F175061755: D46111.id.diff
Wed, Oct 7, 11:59 PM
Unknown Object (File)
Sun, Oct 4, 4:58 AM
Unknown Object (File)
Thu, Oct 1, 2:26 PM
Unknown Object (File)
Thu, Oct 1, 1:35 AM
Unknown Object (File)
Sep 8 2026, 3:54 AM
Unknown Object (File)
Sep 4 2026, 11:19 PM
Unknown Object (File)
Sep 4 2026, 11:05 PM
Unknown Object (File)
Sep 4 2026, 11:04 PM
Subscribers

Details

Reviewers
dougm
Summary

In the past we've had bugs where lack of argument validation allowed a
user-controlled size to be passed to kmem_malloc() and friends.
Explicitly check before rounding up the allocation size to the page size.

This requires a bit of reorganization for kmem_alloc_san(), which wants
both the requested size and the true allocation size.

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Skipped
Unit
Tests Skipped
Build Status
Buildable 58809
Build 55696: arc lint + arc unit

Event Timeline

markj requested review of this revision.Jul 24 2024, 9:31 PM
sys/vm/vm_kern.c
332

addr could be NULL. Add a check.

432

add addr != NULL check.

529

Add addr != NULL check.

alc added inline comments.
sys/vm/vm_kern.c
165

This rounding operation has always been redundant, since vmem_xalloc() also rounds. I would suggest moving the overflow test into vmem_xalloc().

188–191

vmem_xfree() already asserts that the rounded, passed-in size matches the boundary tag, which basically accomplishes the same.