[google/syzkaller] 7465ed: executor: restore dummy_null_handler()

0 views
Skip to first unread message

Alexander Potapenko

unread,
Feb 17, 2026, 4:14:20 AM (6 days ago) Feb 17
to syzk...@googlegroups.com
Branch: refs/heads/gh-readonly-queue/master/pr-6820-d7b7acf8d24bf885bafe1fbf0a979c3ffa59951d
Home: https://github.com/google/syzkaller
Commit: 7465ed05d995335737d4f157ddb344444a102419
https://github.com/google/syzkaller/commit/7465ed05d995335737d4f157ddb344444a102419
Author: Alexander Potapenko <gli...@google.com>
Date: 2026-02-17 (Tue, 17 Feb 2026)

Changed paths:
M executor/common_kvm_amd64_syzos.h

Log Message:
-----------
executor: restore dummy_null_handler()

Somehow one of the previous patches made dummy_null_handler() behave
like uexit_irq_handler(). Restore the original handler behavior.


Commit: 4db8fa928f6610a9c9f808270ed4f6cbdfbded26
https://github.com/google/syzkaller/commit/4db8fa928f6610a9c9f808270ed4f6cbdfbded26
Author: Alexander Potapenko <gli...@google.com>
Date: 2026-02-17 (Tue, 17 Feb 2026)

Changed paths:
M executor/common_kvm_amd64.h

Log Message:
-----------
executor: introduce gpa_to_hva() for x86 SYZOS

L1 guest memory is non-contiguous, but previously host setup assumed
the opposite, using L1 guest addresses as offsets in the host memory
block. This led to subtle bugs in IRQ handling (and possibly elsewhere).

Fix this by using gpa_to_hva() to translate guest physical addresses to
host virtual addresses.
This function is cold, so we can afford O(SYZOS_REGION_COUNT) complexity.


Compare: https://github.com/google/syzkaller/compare/7465ed05d995%5E...4db8fa928f66

To unsubscribe from these emails, change your notification settings at https://github.com/google/syzkaller/settings/notifications

Alexander Potapenko

unread,
Feb 17, 2026, 4:25:04 AM (6 days ago) Feb 17
to syzk...@googlegroups.com
Reply all
Reply to author
Forward
0 new messages