On Tue, 21 Oct 2008 20:08:41 +0400, Sergei Shtylyov <sshtylyov@xxxxxxxxxxxxx> wrote: > > This is the driver for the Toshiba TX4939 SoC ATA controller. > > I'm inclined to ACK the driver (besides, TX49xx patches are > holding up my own series of patches since it needs to modify both > these drivers) but I'm not sure about the error cleanup path now > that I looked at it again -- probably' devres' handles all that > automagically but peering into the sources didn't enlignten me on > how it does it, so I would like to be explicitly assured. :-) Yes, devres do it, as written in Documentation/driver-model/devres.txt. If in doubt, 'probe_failed:' label in drivers/base/dd.c:really_probe() is probably where you want to look at :-) > There are also some nits, mostly ignorable... OK, I will send v6 patch which fixes most of them. > > + tx4939ide_writew(0x0008, base, TX4939IDE_Lo_Burst_Cnt); > > + tx4939ide_writew(0, base, TX4939IDE_Up_Burst_Cnt); > > I think that these fit better to tx4939ide_init_dma(). Unfortunately (and surprisingly) this did not work. The kernel crashed with strange memory corruption. It seems Burst_Cnt must be initialized before any transfer including PIO. I don't know why... > Same question about the error cleanup here -- will the acquired resources > be auto-released? If so, then: > > Acked-by: Sergei Shtylyov <sshtylyov@xxxxxxxxxxxxx> Yes. So I will send v6 patch with your Acked-by line. Thank you very much! --- Atsushi Nemoto
