Rich Ercolani

...huh. Yeah, there it goes, on git (482505fd4). The handwavey parts of the stack trace make me suspicious, let me go try a patch I have... ``` [270108.416880] VERIFY(PageUptodate(pp)) failed...

You changed the title - in experimenting, I only saw it with clones, are you saying you've reproduced it without that?

You'd also probably need a couple patches I've given up on upstreaming to make it not flood the logs with useless messages about lz4 and the crypto code using uninited...

Sure, I wasn't remarking on it as a criticism or something, just to save a round trip of "hey I'm out of space because syslog flooded with these :C"

Someone should poke @behlendorf, he and I looked at this for a few minutes because I had a convenient reproducer set up already, but I forget what came of it...

A note - my limited understanding of the problem is that: * ZFS does a lot of buffering to make `mmap` work on Linux, for reasons I think were partially...

My guess would be it doesn't understand the allocator well enough, but I don't know, I've rarely seen sanitizers do useful things with mmap in particular, so my expectations are...

I believe you're barking up the wrong tree, since that error comes from (I believe) `grub-core/osdep/unix/getroot.c`: ``` if (nvlist_lookup_string (children[i], "path", &device) != 0) error (1, errno, "nvlist_lookup_string (\"path\")"); ```...

Native encryption strikes again. (The versions in `zfs version` differ because you're using the kernel module that shipped with your Ubuntu install and the userland from, I'm going to not-really-guess...

I'm guessing wildly, but what does `zpool get all [pool]` return for the pool(s) on one of those systems?