From: jayakumar alsa Subject: Re: [PATCH 2.6.13.1 1/1] CS5535 AUDIO ALSA driver Date: Tue, 20 Sep 2005 20:03:40 -0400 Message-ID: <47f5dce305092017031a2ba375@mail.gmail.com> References: <200509190639.j8J6dIM4007948@localhost.localdomain> <20050920152830.7ef6733b.akpm@osdl.org> Reply-To: jayakumar.alsa@gmail.com Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Return-path: In-Reply-To: <20050920152830.7ef6733b.akpm@osdl.org> Content-Disposition: inline Sender: alsa-devel-admin@lists.sourceforge.net Errors-To: alsa-devel-admin@lists.sourceforge.net List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , List-Archive: To: Andrew Morton Cc: perex@suse.cz, mj@ucw.cz, alsa-devel@lists.sourceforge.net, linux-kernel@vger.kernel.org List-Id: alsa-devel@alsa-project.org On 9/20/05, Andrew Morton wrote: > By convention we put the asm/ includes after the linux/ includes. Will do. Thanks for the detailed review! >=20 > > + > > +static unsigned short snd_cs5535audio_codec_read(cs5535audio_t *cs5535= au, >=20 > typedefs are unpopular in-kernel. We generally don't get too fussed if a > driver maintainer really wants them there. The main objection is that we > now have two names for the same thing. Plus they cannot be used when > forward-declaring a structure. I just used those typedefs in order to match the style in all the other alsa drivers. For example: % egrep typedef $lintree/sound/pci/*.c als4000.c:typedef struct { atiixp.c:typedef struct snd_atiixp atiixp_t; atiixp.c:typedef struct snd_atiixp_dma atiixp_dma_t; atiixp.c:typedef struct snd_atiixp_dma_ops atiixp_dma_ops_t; atiixp.c:typedef struct atiixp_dma_desc { azt3328.c:typedef struct _snd_azf3328 azf3328_t; azt3328.c:typedef struct azf3328_mixer_reg { bt87x.c:typedef struct snd_bt87x bt87x_t; cmipci.c:typedef struct snd_stru_cmipci cmipci_t; I'm not sure what to do. I'd be happy to take them out. But I woudn't mind leaving them in if that's what alsa convention is. > This isn't right. `timeout' will have a value of -1 if we timed out. > (several instances). How embarrassing. Will fix. :-) > Ditto. Plus we prefer to not put braces around a single statement like t= his. Will fix. > Again. Perhaps a helper function so this code (and its bug) don't get > duplicated so much? Will do. >=20 > We normally put a space after commas. >=20 > Unneeded typecast. Will fix. >=20 > > + dma->index =3D (++(dma->index)) % dma->periods; >=20 > Is no locking needed for dma->index? Looks like I stopped using index after I found I don't need to track the individual descriptors. I'll take it out. >=20 > Just to save a tabstop. Will do. > > + } > > + return IRQ_HANDLED; > > +} >=20 > Perhaps the default case should return IRQ_NONE. In this case I think the code is correct as it is. I check that in the top level irq status register if the irq was from the cs5535 audio. I return NONE if it wasn't there. Then I proceed to read a bunch of other registers, all with the assumption that the irq was from us. I return HANDLED at that point since reading those other registers has cleared any irq from us even if it was from an unexpected src. > > + cs5535au =3D kcalloc(1, sizeof(*cs5535au), GFP_KERNEL); >=20 > We have kzalloc() now. Yes, Takashi pointed that out too. But I did the patch against 13.1 which doesn't have kzalloc. I guess I'll redo the patch against -mm and switch to kzalloc. > Please consider reworking functions such as the above so as to have a > single `reutrn' statement. Or one `return' for success and just one for > the error paths. Reason: a) it's easier to chack that all resources are > being freed on the error paths and b) It's easier to add new stuff which > allocates new resources which need to be released on error paths. Involv= es > goto spaghetti. Will do. >=20 > The handling of `dev' is racy ;) Ok. Will look at that. >=20 > > + addr =3D (u32)substream->runtime->dma_addr; >=20 > Nope, _snd_pcm_runtime.addr has type dma_addr_t, which is an opaque type, > 64-bit on some platforms. I expect this driver will blow up on those > platforms. The 5535 hw's dma descriptor is only 32 bit capable. I guess I should look into informing the dma alloc that the buffers need to be in the lower 32. Would it be okay to drop the upper then? >From a practical standpoint, the 5535 is only used in x86-32 embeddeded systems so high mem probably won't occur. But you are right, I'll go and try make it right. > +#define cs_writel(reg, val) outl(val, (int) cs5535au->port + reg) > +#define cs_writeb(reg, val) outb(val, (int) cs5535au->port + reg) > +#define cs_readl(reg) inl((unsigned short) (cs5535au->port + re= g)) > +#define cs_readw(reg) inw((unsigned short) (cs5535au->port + re= g)) > +#define cs_readb(reg) inb((unsigned short) (cs5535au->port + re= g)) >=20 > erk. subsystem-wide helper macros which reference local variables? Ok. I'll change that. Thanks, jk ------------------------------------------------------- SF.Net email is sponsored by: Tame your development challenges with Apache's Geronimo App Server. Download it for free - -and be entered to win a 42" plasma tv or your very own Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.php