ATA_DMA errors (and fs corruption!)

[Martin]

twesky wrote:
> I am having ATA_DMA errors on 5.4R and 5 STABLE up to June 16 (haven't
> done a cvsup again).  It doesn't happen on 5.3R or lower.

I have got same problem. I tried yesterday's kernel and I got lots of
ATA DMA errors. A question: do you have a VIA IDE controller like mine?

atapci0: <VIA 8235 UDMA133 controller> port
0xfc00-0xfc0f,0x376,0x170-0x177,0x3f6,0x1f0-0x1f7 at device 17.1 on pci0

atapci0 at pci0:17:1:      class=0x01018a card=0x05711849 chip=0x05711106
rev=0x06 hdr=0x00
    vendor   = 'VIA Technologies Inc'
    device   = 'VT82xxxx EIDE Controller (All VIA Chipsets)'
    class    = mass storage
    subclass = ATA

Today, I noticed, the short experiment with the latest -STABLE destroyed
a part of my /usr partition. It looked like this (with May 9th kernel
today):

kernel: handle_workitem_freeblocks: block count
kernel: bad block 50333952, ino 1743780
kernel: pid 56 (syncer), uid 0 inumber 1743780 on /usr: bad block
kernel: bad block 3221252091, ino 1743780
klotz kernel: pid 56 (syncer), uid 0 inumber 1743780 on /usr: bad block
kernel: bad block 144119931884736777, ino 1743780
kernel: pid 56 (syncer), uid 0 inumber 1743780 on /usr: bad block
kernel: bad block 72340173158093844, ino 1743780
kernel: pid 56 (syncer), uid 0 inumber 1743780 on /usr: bad block
kernel: bad block 1104111992832, ino 1743780
kernel: pid 56 (syncer), uid 0 inumber 1743780 on /usr: bad block
kernel: handle_workitem_freeblocks: block count
kernel: handle_workitem_freeblocks: block count
kernel: bad block 1865342872522620032, ino 1743783

While shutting down I got this:

Jun 19 22:04:21 klotz kernel: /usr: unmount pending error: blocks
-3561100369582 68157 files 0

I restored the fs in single-user mode. And now it runs fine with
the kernel (May 9th).

See also my earlier post.

Martin