Mark Furneaux

That sounds like a great test. I appreciate your suggestion. Despite my use of ZFS for almost a decade I am not very familiar with the internals. I will reproduce...

I tested the existing (non-debug) kernel and set `ZFS_DEBUG_MODIFY` and no abnormal messages were emitted in `dbgmsg`. I compiled `master` with `--enable-debug` and `gcc` (`cc1`) triggered this panic: ``` Jun...

My apologies, my testing did show it only with memory mapped files, but I have not definitively reproduced without a clone. I was just trying to make the title less...

I'll try and do a test with those options some time this week.

Good to note. I run my own kernel build so it shouldn't be an issue to get the sanitizer support built-in on my test bench.

I compiled 5.18.14 with `CONFIG_KASAN=y` and `CONFIG_DEBUG_INFO=y`, and configured ZFS 2.1.5 release (with the patches from https://github.com/openzfs/zfs/pull/13208) with `--enable-asan --enable-systemd --enable-debug --enable-debuginfo`. After over an hour I did get another...

Correct me if I am wrong, but are these Linux config options not what is required? Are there more options I am missing? ``` CONFIG_KASAN=y CONFIG_KASAN_GENERIC=y CONFIG_KASAN_OUTLINE=y CONFIG_KASAN_STACK=y ``` I...

Turns out I also had that enabled as well (I configured this kernel weeks ago and just got to running it now): ``` CONFIG_UBSAN=y CONFIG_UBSAN_BOUNDS=y CONFIG_UBSAN_ONLY_BOUNDS=y CONFIG_UBSAN_SHIFT=y CONFIG_UBSAN_BOOL=y CONFIG_UBSAN_ENUM=y CONFIG_UBSAN_SANITIZE_ALL=y...

Some more test results. I re-tested with `CONFIG_KASAN_VMALLOC=y` and ZFS mainline as of ab49df487b5dfee8d045195c30d3bf1b00c23910 and the results were the same - no output from any sanitizer, trace as follows: ```...

I'd be surprised if the buffer is zeroed before being filled in. In the one case I managed to see the page contents it was definitely not zeroed so that...