From: Johan Hovold <johan@kernel.org> To: Laurentiu Palcu <laurentiu.palcu@intel.com> Cc: Johan Hovold <johan@kernel.org>, Mark Brown <broonie@kernel.org>, Samuel Ortiz <sameo@linux.intel.com>, Lee Jones <lee.jones@linaro.org>, Octavian Purdila <octavian.purdila@intel.com>, linux-spi@vger.kernel.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH v3 1/2] spi: add support for DLN-2 USB-SPI adapter Date: Wed, 19 Nov 2014 11:09:32 +0100 [thread overview] Message-ID: <20141119100932.GC7857@localhost> (raw) In-Reply-To: <20141119095343.GD2583@lpalcu-linux> On Wed, Nov 19, 2014 at 11:53:44AM +0200, Laurentiu Palcu wrote: > On Tue, Nov 18, 2014 at 06:06:27PM +0100, Johan Hovold wrote: > > On Tue, Nov 18, 2014 at 02:09:58PM +0200, Laurentiu Palcu wrote: > > > > > +/* > > > + * Copy the data to DLN2 buffer and change the byte order to LE, requested by > > > + * DLN2 module. SPI core makes sure that the data length is a multiple of word > > > + * size. > > > + */ > > > +static int dln2_spi_copy_to_buf(u8 *dln2_buf, const u8 *src, u16 len, u8 bpw) > > > +{ > > > +#ifdef __LITTLE_ENDIAN > > > + memcpy(dln2_buf, src, len); > > > +#else > > > + if (bpw <= 8) { > > > + memcpy(dln2_buf, src, len); > > > + } else if (bpw <= 16) { > > > + __le16 *d = (__le16 *)dln2_buf; > > > + u16 *s = (u16 *)src; > > > + > > > + len = len / 2; > > > + while (len--) > > > + put_unaligned_le16(get_unaligned(s++), d++); > > > > You said the dln2 buffer was properly aligned, right? Then you don't > > need to use put_unaligned_lexx here... > > I know, and I was very close not to use it, but, as you said, others > might not be as careful That was in a different context. :) > and if Diolan FW changes anything in the > header's structure in the future, the alignment will be off. This is > just a precaution. I know it's a little bit costly but this will affect > only BE machines and only if bits-per-word is bigger than 8. I don't think you should worry about hypothetical changes to the protocol. You verified the that headers are a multiple of four, so no need for the unaligned macros. > > > + } else { > > > + __le32 *d = (__le32 *)dln2_buf; > > > + u32 *s = (u32 *)src; > > > + > > > + len = len / 4; > > > + while (len--) > > > + put_unaligned_le32(get_unaligned(s++), d++); > > > + } > > > +#endif > > > + > > > + return 0; > > > +} > > > + > > > +/* > > > + * Copy the data from DLN2 buffer and convert to CPU byte order since the DLN2 > > > + * buffer is LE ordered. SPI core makes sure that the data length is a multiple > > > + * of word size. > > > + */ > > > +static int dln2_spi_copy_from_buf(u8 *dest, const u8 *dln2_buf, u16 len, u8 bpw) > > > +{ > > > +#ifdef __LITTLE_ENDIAN > > > + memcpy(dest, dln2_buf, len); > > > +#else > > > + if (bpw <= 8) { > > > + memcpy(dest, dln2_buf, len); > > > + } else if (bpw <= 16) { > > > + u16 *d = (u16 *)dest; > > > + __le16 *s = (__le16 *)dln2_buf; > > > + > > > + len = len / 2; > > > + while (len--) > > > + put_unaligned(get_unaligned_le16(s++), d++) > > > > ...or get_unaligned_lexx here. > > > > > + } else { > > > + u32 *d = (u32 *)dest; > > > + __le32 *s = (__le32 *)dln2_buf; > > > + > > > + len = len / 4; > > > + while (len--) > > > + put_unaligned(get_unaligned_le32(s++), d++) > > > + } > > > +#endif > > > + > > > + return 0; > > > +} > > > > Did you check the alignment of the SPI buffers as well? I'd assume they > > were DMA-able and thus properly aligned, and then you do not need to use > > any unaligned helpers above at all. > > I did... The spi_transfer struct documentation says that those tx_buf > and rx_buf buffers are dma-safe but, apparently, I found drivers that do > not really use alligned buffers for transfers. Look, for example, at > max1111.c. Those buffers don't look aligned to me... That looks like a bug to me. DMA-buffers should never be allocated as part of a larger structure. > Well, this driver > is probably not the best example since BPW is 8, but you got my point. > Other drivers, indeed, use the ____cacheline_aligned attribute and > should be dma-safe. I think you can remove the unaligned macros altogether. Johan
WARNING: multiple messages have this Message-ID (diff)
From: Johan Hovold <johan-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org> To: Laurentiu Palcu <laurentiu.palcu-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> Cc: Johan Hovold <johan-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>, Mark Brown <broonie-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>, Samuel Ortiz <sameo-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>, Lee Jones <lee.jones-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>, Octavian Purdila <octavian.purdila-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>, linux-spi-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org Subject: Re: [PATCH v3 1/2] spi: add support for DLN-2 USB-SPI adapter Date: Wed, 19 Nov 2014 11:09:32 +0100 [thread overview] Message-ID: <20141119100932.GC7857@localhost> (raw) In-Reply-To: <20141119095343.GD2583@lpalcu-linux> On Wed, Nov 19, 2014 at 11:53:44AM +0200, Laurentiu Palcu wrote: > On Tue, Nov 18, 2014 at 06:06:27PM +0100, Johan Hovold wrote: > > On Tue, Nov 18, 2014 at 02:09:58PM +0200, Laurentiu Palcu wrote: > > > > > +/* > > > + * Copy the data to DLN2 buffer and change the byte order to LE, requested by > > > + * DLN2 module. SPI core makes sure that the data length is a multiple of word > > > + * size. > > > + */ > > > +static int dln2_spi_copy_to_buf(u8 *dln2_buf, const u8 *src, u16 len, u8 bpw) > > > +{ > > > +#ifdef __LITTLE_ENDIAN > > > + memcpy(dln2_buf, src, len); > > > +#else > > > + if (bpw <= 8) { > > > + memcpy(dln2_buf, src, len); > > > + } else if (bpw <= 16) { > > > + __le16 *d = (__le16 *)dln2_buf; > > > + u16 *s = (u16 *)src; > > > + > > > + len = len / 2; > > > + while (len--) > > > + put_unaligned_le16(get_unaligned(s++), d++); > > > > You said the dln2 buffer was properly aligned, right? Then you don't > > need to use put_unaligned_lexx here... > > I know, and I was very close not to use it, but, as you said, others > might not be as careful That was in a different context. :) > and if Diolan FW changes anything in the > header's structure in the future, the alignment will be off. This is > just a precaution. I know it's a little bit costly but this will affect > only BE machines and only if bits-per-word is bigger than 8. I don't think you should worry about hypothetical changes to the protocol. You verified the that headers are a multiple of four, so no need for the unaligned macros. > > > + } else { > > > + __le32 *d = (__le32 *)dln2_buf; > > > + u32 *s = (u32 *)src; > > > + > > > + len = len / 4; > > > + while (len--) > > > + put_unaligned_le32(get_unaligned(s++), d++); > > > + } > > > +#endif > > > + > > > + return 0; > > > +} > > > + > > > +/* > > > + * Copy the data from DLN2 buffer and convert to CPU byte order since the DLN2 > > > + * buffer is LE ordered. SPI core makes sure that the data length is a multiple > > > + * of word size. > > > + */ > > > +static int dln2_spi_copy_from_buf(u8 *dest, const u8 *dln2_buf, u16 len, u8 bpw) > > > +{ > > > +#ifdef __LITTLE_ENDIAN > > > + memcpy(dest, dln2_buf, len); > > > +#else > > > + if (bpw <= 8) { > > > + memcpy(dest, dln2_buf, len); > > > + } else if (bpw <= 16) { > > > + u16 *d = (u16 *)dest; > > > + __le16 *s = (__le16 *)dln2_buf; > > > + > > > + len = len / 2; > > > + while (len--) > > > + put_unaligned(get_unaligned_le16(s++), d++) > > > > ...or get_unaligned_lexx here. > > > > > + } else { > > > + u32 *d = (u32 *)dest; > > > + __le32 *s = (__le32 *)dln2_buf; > > > + > > > + len = len / 4; > > > + while (len--) > > > + put_unaligned(get_unaligned_le32(s++), d++) > > > + } > > > +#endif > > > + > > > + return 0; > > > +} > > > > Did you check the alignment of the SPI buffers as well? I'd assume they > > were DMA-able and thus properly aligned, and then you do not need to use > > any unaligned helpers above at all. > > I did... The spi_transfer struct documentation says that those tx_buf > and rx_buf buffers are dma-safe but, apparently, I found drivers that do > not really use alligned buffers for transfers. Look, for example, at > max1111.c. Those buffers don't look aligned to me... That looks like a bug to me. DMA-buffers should never be allocated as part of a larger structure. > Well, this driver > is probably not the best example since BPW is 8, but you got my point. > Other drivers, indeed, use the ____cacheline_aligned attribute and > should be dma-safe. I think you can remove the unaligned macros altogether. Johan -- To unsubscribe from this list: send the line "unsubscribe linux-spi" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2014-11-19 10:09 UTC|newest] Thread overview: 48+ messages / expand[flat|nested] mbox.gz Atom feed top 2014-11-18 12:09 [PATCH v3 0/2] Add SPI support for Diolan DLN2 Laurentiu Palcu 2014-11-18 12:09 ` Laurentiu Palcu 2014-11-18 12:09 ` [PATCH v3 1/2] spi: add support for DLN-2 USB-SPI adapter Laurentiu Palcu 2014-11-18 17:06 ` Johan Hovold 2014-11-18 17:06 ` Johan Hovold 2014-11-19 9:53 ` Laurentiu Palcu 2014-11-19 10:09 ` Johan Hovold [this message] 2014-11-19 10:09 ` Johan Hovold 2014-11-19 10:47 ` [PATCH v4 " Laurentiu Palcu 2014-11-19 10:47 ` Laurentiu Palcu 2014-11-19 12:21 ` Johan Hovold 2014-11-19 12:21 ` Johan Hovold 2014-11-19 13:07 ` [PATCH v5 " Laurentiu Palcu 2014-11-19 13:07 ` Laurentiu Palcu 2014-11-19 13:51 ` Johan Hovold 2014-11-19 13:51 ` Johan Hovold 2014-11-19 14:12 ` Laurentiu Palcu 2014-11-19 14:12 ` Laurentiu Palcu 2014-11-26 10:09 ` [PATCH v6 " Laurentiu Palcu 2014-11-26 18:25 ` Mark Brown 2014-11-26 18:25 ` Mark Brown 2014-11-27 10:30 ` Laurentiu Palcu 2014-11-27 10:30 ` Laurentiu Palcu 2014-11-28 13:09 ` [PATCH v7 " Laurentiu Palcu 2014-11-28 14:10 ` Mark Brown 2014-11-28 14:57 ` Laurentiu Palcu 2014-11-28 14:57 ` Laurentiu Palcu 2014-11-28 15:14 ` Mark Brown 2014-11-28 15:14 ` Mark Brown 2014-11-28 15:31 ` Laurentiu Palcu 2014-11-28 15:31 ` Laurentiu Palcu 2014-11-28 16:01 ` Mark Brown 2014-11-28 18:13 ` Laurentiu Palcu 2014-11-28 18:13 ` Laurentiu Palcu 2014-11-28 18:35 ` Mark Brown 2014-11-28 18:35 ` Mark Brown 2014-11-28 19:27 ` Laurentiu Palcu 2014-11-28 19:27 ` Laurentiu Palcu 2014-11-28 19:46 ` Mark Brown 2014-11-28 19:46 ` Mark Brown 2014-11-28 19:21 ` Octavian Purdila 2014-11-28 19:21 ` Octavian Purdila 2014-12-05 18:34 ` Mark Brown 2014-12-05 18:34 ` Mark Brown 2014-11-18 12:09 ` [PATCH v3 2/2] mfd: dln2: add support for USB-SPI module Laurentiu Palcu 2014-11-18 12:09 ` Laurentiu Palcu 2014-11-19 16:29 ` Lee Jones 2014-11-19 16:29 ` Lee Jones
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20141119100932.GC7857@localhost \ --to=johan@kernel.org \ --cc=broonie@kernel.org \ --cc=laurentiu.palcu@intel.com \ --cc=lee.jones@linaro.org \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-spi@vger.kernel.org \ --cc=octavian.purdila@intel.com \ --cc=sameo@linux.intel.com \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.