Re: [Xen-users] 3w-xxxx / 3ware 8006-2LP corruption issues using Xen kernel

[Bas Verhoeven <xen@xxxxxxxxxx>]

Holm Kapschitzki wrote:
> Bas Verhoeven schrieb:
Hi Holm,
> look at my ealier post, i describe the same problem.
>
> http://www.nabble.com/dom0---tar:-Skipping-to-next-header-td16558409.html
>
> so i get the error with etch 32 bit / 64 bit, xen 3.1 / 3.2 , with 
> 2.6.18 kernel xen , with gentoo kernel 2.6.20r6 xen. It wasnt all the 
> time. But i have to reboot the maschine to get it solved for a while. 
> So i testet it with ca. 5 machines, setup in different ways with other 
> kernels.
In a way I'm happy I'm not the only one experiencing this problem. Are 
you using the exact same controller as I am?
I did experience some issues when I would remove most of the memory; so 
the system would be left with 1GB of memory. At that point, running my 
script would cause several errors, ending up in the partition becoming 
read-only:
   PCI-DMA: Out of SW-IOMMU space for 65536 bytes at device 0000:02:01.0
   3w-xxxx: tw_map_scsi_sg_data(): pci_map_sg() failed.
   PCI-DMA: Out of SW-IOMMU space for 65536 bytes at device 0000:02:01.0
   3w-xxxx: tw_map_scsi_sg_data(): pci_map_sg() failed.
   ...
   sd 0:0:0:0: SCSI error: return code = 0x00070000
   end_request: I/O error, dev sda, sector 3068774
   Buffer I/O error on device dm-0, logical block 321289
   lost page write due to I/O error on dm-0
   Buffer I/O error on device dm-0, logical block 321290
   lost page write due to I/O error on dm-0
   ...
   end_request: I/O error, dev sda, sector 11794406
   Aborting journal on device dm-0.
   ext3_abort called.
   EXT3-fs error (device dm-0): ext3_journal_start_sb: Detected aborted
   journal
   Remounting filesystem read-only
   __journal_remove_journal_head: freeing b_frozen_data


This problems seems to be unrelated tho, and some googling pointed me to 
some 'swiotlb' kernel parameter 
(https://bugzilla.novell.com/show_bug.cgi?id=299641), which I set to 32M 
and seems to run OK for now.
Data is still being written corrupted to disk tho.
> I think i could be a kernel compile parameter? or via chipset in 
> relation to 3ware raid controller?
Well, I hardly doubt it's something in the hardware itself. That just 
does not explain why everything works fine under a non-Xen kernel.
All kernels I tried have the 3ware driver loaded as a module. The 
drivers under both kernels appear to be the same:
   p-dom0:/usr/src/xen-3.2.0/linux-2.6.18-xen.hg# sha1sum
   drivers/scsi/3w-x*
   d9da8960f6e98b783b4893cde51a303d97ce98d8  drivers/scsi/3w-xxxx.c
   2610261f86b4eb05a5d08c1f90f09410f1eb7c98  drivers/scsi/3w-xxxx.h

   p-dom0:/usr/src/linux-2.6.18.8# sha1sum drivers/scsi/3w-x*
   d9da8960f6e98b783b4893cde51a303d97ce98d8  drivers/scsi/3w-xxxx.c
   2610261f86b4eb05a5d08c1f90f09410f1eb7c98  drivers/scsi/3w-xxxx.h

So whatever is breaking stuff, must be something in the Xen code? I'm 
going to compile a kernel with the 3w-xxxx driver compiled in, but I 
doubt that helps.
Is there even anyone that uses the same controller and has no problems 
at all?
Cheers,

Bas Verhoeven

> Greets Holm
>
>
> _______________________________________________
> Xen-users mailing list
> Xen-users@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-users

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users