[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] vbd backend dropping block I/Os?


I have a block device (implemented as a kernel module)
that we've written for our own purposes which we've
attempted to load as an IDD...

I have done the following:
1. added the appropriate [ phy:/dev/somedevice,/dev/somedevice,w ]
entry to the guest domU config
2. issued a block-based mknod /dev/somedevice
3. Finally: I sent some blocks to the device with dd from
a loaded guest on the blkfront side. I have debug printk's
in the device that periodically let me know how many BIOs
the block device has received on the blkback side: but the
number of requests received by the device does not match
the number of requests sent to it from the guest kernel
through the device channel. (It's about a 10 to 1 ratio
of missing requests) -> yet the dd command seems to think
they were all sent successfully from the guest's perspective.

I've done a 'xenstore-ls' and verified that the sector
size of the device that xen perceives is identical to
that of my block device.

Is there any obvious reason why this might happen?

Your help is much appreciated.

- Michael

Attachment: signature.asc
Description: OpenPGP digital signature

Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.