On Mon, 20 Apr 2020 at 12:38, Peter Maydell <peter.maydell@xxxxxxxxxx> wrote: > Whoops, good point. I was testing this via lkvm, so it's > actually using a 9p filesystem... I'll see if I can figure > out how to test with an ext3 fs, which I think is the one > we most care about. After some effort wrestling with kvmtool (which assumes that if you provide it a disk image then you must have wanted that to be your rootfs and can only be persuaded otherwise via some undocumented and arcane options), I did a test with ext4: bash-4.4# /qemu-no-fix /readdir-bug dir=0x76108 readdir(dir)=(nil) errno=75: Value too large for defined data type bash-4.4# /qemu-fixed /readdir-bug dir=0x76108 readdir(dir)=0x76128 errno=0: Success (where the host kernel has Linus' fcntl patch, qemu-no-fix is a current-git-master QEMU and qemu-fixed is one with the patch in my previous email). So for Linus' patch: Tested-by: Peter Maydell <peter.maydell@xxxxxxxxxx> If 9pfs could be persuaded to honour the fcntl flag too that would be really nice. thanks -- PMM
