1 files changed, 12 insertions, 0 deletions
diff --git a/results/classifier/accel-gemma3:12b/tcg/2149 b/results/classifier/accel-gemma3:12b/tcg/2149
new file mode 100644
index 000000000..194155b3c
--- /dev/null
+++ b/results/classifier/accel-gemma3:12b/tcg/2149
@@ -0,0 +1,12 @@
+
+Segfault in libvhost-user and libvduse because of invalid pointer arithmetic with indirect read
+Description of problem:
+Hello, this is my first experience communicating with open-source community. I have already reported the problem and have submitted patches through qemu-devel mailing list https://mail.gnu.org/archive/html/qemu-devel/2024-01/msg02533.html, as instructed in https://www.qemu.org/docs/master/devel/submitting-a-patch.html, albeit getting no response from any maintainer. I know, that everyone are very busy and are spammed everyday from millions of threads, but I am getting very upset, that such a trivial bug lives in code base for many years and even have been copied to "sister"-library without proper review. So, excuse me, if I am taking this issue too personally.
+
+The problem - when one tries to use libvhost-user\libvduse and triggers for some reason non-zero-copy mode (like pushing a lot of data) of indirect descriptor reading routine `virtqueue_read_indirect_desc`, any time one got to read more than one descriptor - one would overwrite stack and depending on one's luck getting some weird behaviour, or simple crash moments later, when other code tries to access broken data.
+
+Steps to reproduce are non-trivial, because depends on one's host and VM (one simply gets random crashes here and there, with core dumps pointing somewhere around given libraries), but anyone who can read C code, can clearly see that pointer arithmetic of `struct vring_desc *desc` is wrong.
+
+Maybe, I got instructions wrong and posted fixes to wrong mailing list, maybe, nobody cares, so thank you for attention. I'll be glad to hear any advice on how can I help with fixing this simple error, besides what has been done already.
+
+Thank you.