On Tue, 2005-01-11 at 11:26 -0500, goggin, edward wrote: > Looks like domap() in multipath/main.c is not resuming a mapped device > before removing the > device in the event of an error returned from dm_addmap(). This results in > leaving the kernel > block_device structure for the mapped device remaining "frozen", so that > subsequent > attempts to reload corrected configuration data hang on the bd_mount_sem of > the block_device > since the instance of the mapped device structure holding the DMF_FS_LOCKED > flag bit > had been destroyed. > Indeed, this code is unsafe. The fix is amazingly simple, as the suspend / resume cycle is not even necessary. I removed it for the next release. regards, -- christophe varoqui <christophe.varoqui@xxxxxxx>
