This series adds support for device change notifications to the pnfs core and block layout driver. To properly implement NOTIFY_DEVICEID4_CHANGE I had to move GETDEVICEINFO decoding into kernel space, and implement my own trivial striping and concat layers in the pnfs blocklayout client. Why? DM and MD are really, really unhappy about multiple raid devices using the same physical devices. Which is fairly understandable from their point of view, given that they normally manage on-disk metadata and would rather do that exclusively. But we get all our metadata from the server and couldn't care less about their problems, we have our own problem called deviceid management.. Once we get a NOTIFY_DEVICEID4_CHANGE callback we need to stop new I/O using the existing mapping, so we have to drop it from the deviceid cache. At the same time another layout could easily request the same deviceid again and for some period we'd have multiple volume structures pointing to the same device. Note that we do not break userspace compatibility - we still use blkmapd to translate signatures to device numbers, and to do so we re-encode the SIMPLE volumes we need translated to XDR and send it up to blkmapd just for that. This work was sponsored by NetApp, Inc. Changes since V1: - support CB_NOTIFY_DEVICEID unconditionally -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
