[Linux-cachefs] vmware on an nfs export backed by fs-cache oopses

[Shaya Potter]

Shaya Potter wrote:
> So as an experiment, I was trying to run a vmware vm stored on an nfs 
> server w/ a fs-cache backing store (to make it easier on the server, 
> especially after a reboot).
> 
> Anyways, it doesn't work, at least w/ vmplayer, it reliably oopses in 
> the same manner.
> 
> this is on 2.6.17 with my patched kernel from yesterday (from steved's 
> html dir)
> 
> BUG: unable to handle kernel NULL pointer dereference at virtual address 
> 00000001
>  printing eip:
> 00000001
> *pde = 00000000
> Oops: 0000 [#1]
> PREEMPT SMP
> Modules linked in: ipv6 sworks_agp nvidia_agp ati_agp amd_k7_agp ali_agp 
> vmnet parport_pc parport vmmon ext3 jbd dm_mod fuse tsdev ipw2200 
> ieee80211 ieee80211_crypt sdhci mmc_core snd_intel8x0 8250_pci 8250 
> serial_core snd_pcm_oss snd_mixer_oss snd_intel8x0m snd_ac97_codec 
> snd_ac97_bus snd_pcm snd_timer snd irtty_sir sir_dev nsc_ircc irda 
> crc_ccitt i2c_i801 i2c_core soundcore snd_page_alloc intel_agp agpgart 
> evdev usbhid pcmcia firmware_class yenta_socket rsrc_nonstatic 
> pcmcia_core video thermal processor ibm_acpi fan container button 
> battery ac unionfs sbp2 ohci1394 ieee1394 usb_storage ohci_hcd uhci_hcd 
> ehci_hcd usbcore
> CPU:    0
> EIP:    0060:[<00000001>]    Tainted: P      VLI
> EFLAGS: 00213246   (2.6.17 #8)
> EIP is at 0x1
> eax: e039ad40   ebx: c120942c   ecx: 000000ff   edx: 00000000
> esi: d04a2c24   edi: d04a1ea0   ebp: c1105500   esp: c83d9ee0
> ds: 007b   es: 007b   ss: 0068
> Process vmware-vmx (pid: 5956, threadinfo=c83d8000 task=d014ea90)
> Stack: <0>c015769a d04a2c24 c1105500 00000400 00000000 b7fa8000 d04a2c24 
> dcf5e200
>        df1dbdcc d0474ce4 00000000 c120942c 00000001 082a8025 d04a1ea0 
> c0157d2b
>        d04a1ea0 d0c2cb7c c120942c 082a8025 c0158d07 dd1e3d7c 00000ea0 
> d0c2cb7c
> Call Trace:
>  <c015769a> do_wp_page+0x2c6/0x3a8   <c0157d2b> 
> __handle_mm_fault+0x5af/0x868
>  <c0158d07> vma_link+0x93/0xe0   <c011833b> do_page_fault+0x127/0x634
>  <c0118214> do_page_fault+0x0/0x634   <c0103e8f> error_code+0x4f/0x54
> Code:  Bad EIP value.
> EIP: [<00000001>] 0x1 SS:ESP 0068:c83d9ee0

hmm, this might not be a fs-cache issue, as seems to give similiar 
results with regular NFS (albiet patched kernel), I'll have to try it 
without it.