Re: [PATCH v2] rbtree: fix the red root

[Qian Cai <cai@lca.pw>]

On 1/16/19 9:37 AM, Esme wrote:
> I have been off but back now, I had fetch'd current again and the diagnostics look a bit different, maybe I just got lucky.  Instead of fork'ng the test case (which is fairly aggressive in any case), interacting from the serial port with sig-int ^C tend's to trigger enough to hit something.  I'll get the page_owner sorted soon.
> 
> How I'm running;
> 
> qemu-system-x86_64 -kernel /home/files/dl/linux//arch/x86/boot/bzImage -append console=ttyS0 root=/dev/sda debug earlyprintk=serial slub_debug=QUZFP page_owner=on -hda stretch.img -net user,hostfwd=tcp::10021-:22 -net nic -enable-kvm -nographic -m 2G -smp 2
> 
> It's somewhat random I guess that in the last two CPU context dump's printed out, we see RAX and CR2 off by 4 from one another.
> 
> root@syzkaller:~# gcc -o test3 test3.c
> [  392.754148] ata1: lost interrupt (Status 0x50)
> [  392.754478] ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen
> [  392.759687] ata1.00: failed command: READ DMA
> [  392.761902] ata1.00: cmd c8/00:86:00:00:00/00:00:00:00:00/e0 tag 0 dma 68608 out
> [  392.761902]          res 40/00:01:00:00:00/00:00:00:00:00/a0 Emask 0x4 (timeout)
> [  392.768541] ata1.00: status: { DRDY }
> [  392.769532] ata1: soft resetting link
> [  392.937942] ata1.00: configured for MWDMA2
> [  392.945624] ata1: EH complete

While you are gathering page_owner (or kdump), it might be useful to use virtio
storage driver instead of legacy IDE here, as looks like this ATA was busted.