On 01/01/10 11:58, Quentin Garnier wrote: [---] > Having said that, you might question the utility of a static instance > for a pseudo-device, but maybe someone will have a use for that. I'll leave that part as it is for now, and revisit it later. But there's another problem with the patch. To summarize, when a cgd device is being unconfigured, disk_destroy() is called from cgd_ioctl_clr(). This destroys the "open" and "raw" locks. Once the clr ioctl is done, cgdclose() is called, which calls dk_close(), which attempts to acquire the destroyed "open" lock, causing a panic. As far as I have gathered, the cf-stuff is on a higher level of abstraction than the dk-stuff (i.e. initialized earlier, and terminated later). If I've understood this concept correctly, would it make sense to move disk_destroy() to cgd_detach()? (I notice that disk_init() is in cgd_attach()). -- Kind regards, Jan Danielsson
Attachment:
signature.asc
Description: OpenPGP digital signature