From: Adrian Hunter <adrian.hunter@intel.com>
To: Russell King <rmk+kernel@arm.linux.org.uk>,
Ulf Hansson <ulf.hansson@linaro.org>
Cc: Marcin Wojtas <mw@semihalf.com>,
Gregory CLEMENT <gregory.clement@free-electrons.com>,
Shawn Guo <shawnguo@kernel.org>,
Sascha Hauer <kernel@pengutronix.de>,
linux-mmc@vger.kernel.org
Subject: Re: [PATCH v2 08/24] mmc: sdhci: avoid unnecessary mapping/unmapping of align buffer
Date: Tue, 29 Dec 2015 15:44:38 +0200 [thread overview]
Message-ID: <56828E46.5090600@intel.com> (raw)
In-Reply-To: <E1aAypq-00016e-Lv@rmk-PC.arm.linux.org.uk>
On 21/12/15 13:41, Russell King wrote:
> Unnecessarily mapping and unmapping the align buffer for SD cards is
> expensive: performance measurements on iMX6 show that this gives a hit
> of 10% on hdparm buffered disk reads.
>
> MMC/SD card IO comes from the mm/vfs which gives us page based IO, so
> for this case, the align buffer is not going to be used. However, we
> still map and unmap this buffer.
>
> Eliminate this by switching the align buffer to be a DMA coherent
> buffer, which needs no DMA maintenance to access the buffer.
Did you consider putting the align buffer in the same allocation
as the adma_table?
>
> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
> ---
> drivers/mmc/host/sdhci.c | 53 ++++++++++++++++--------------------------------
> 1 file changed, 18 insertions(+), 35 deletions(-)
>
> diff --git a/drivers/mmc/host/sdhci.c b/drivers/mmc/host/sdhci.c
> index 3e718e465a1b..8a4eb1c41f3e 100644
> --- a/drivers/mmc/host/sdhci.c
> +++ b/drivers/mmc/host/sdhci.c
> @@ -465,8 +465,6 @@ static void sdhci_adma_mark_end(void *desc)
> static int sdhci_adma_table_pre(struct sdhci_host *host,
> struct mmc_data *data)
> {
> - int direction;
> -
> void *desc;
> void *align;
> dma_addr_t addr;
> @@ -483,20 +481,9 @@ static int sdhci_adma_table_pre(struct sdhci_host *host,
> * We currently guess that it is LE.
> */
>
> - if (data->flags & MMC_DATA_READ)
> - direction = DMA_FROM_DEVICE;
> - else
> - direction = DMA_TO_DEVICE;
> -
> - host->align_addr = dma_map_single(mmc_dev(host->mmc),
> - host->align_buffer, host->align_buffer_sz, direction);
> - if (dma_mapping_error(mmc_dev(host->mmc), host->align_addr))
> - goto fail;
> - BUG_ON(host->align_addr & host->align_mask);
> -
> host->sg_count = sdhci_pre_dma_transfer(host, data);
> if (host->sg_count < 0)
> - goto unmap_align;
> + return -EINVAL;
>
> desc = host->adma_table;
> align = host->align_buffer;
> @@ -567,22 +554,7 @@ static int sdhci_adma_table_pre(struct sdhci_host *host,
> /* nop, end, valid */
> sdhci_adma_write_desc(host, desc, 0, 0, ADMA2_NOP_END_VALID);
> }
> -
> - /*
> - * Resync align buffer as we might have changed it.
> - */
> - if (data->flags & MMC_DATA_WRITE) {
> - dma_sync_single_for_device(mmc_dev(host->mmc),
> - host->align_addr, host->align_buffer_sz, direction);
> - }
> -
> return 0;
> -
> -unmap_align:
> - dma_unmap_single(mmc_dev(host->mmc), host->align_addr,
> - host->align_buffer_sz, direction);
> -fail:
> - return -EINVAL;
> }
>
> static void sdhci_adma_table_post(struct sdhci_host *host,
> @@ -602,9 +574,6 @@ static void sdhci_adma_table_post(struct sdhci_host *host,
> else
> direction = DMA_TO_DEVICE;
>
> - dma_unmap_single(mmc_dev(host->mmc), host->align_addr,
> - host->align_buffer_sz, direction);
> -
> /* Do a quick scan of the SG list for any unaligned mappings */
> has_unaligned = false;
> for_each_sg(data->sg, sg, host->sg_count, i)
> @@ -3003,6 +2972,10 @@ int sdhci_add_host(struct sdhci_host *host)
> host->adma_table_sz,
> &host->adma_addr,
> GFP_KERNEL);
> + host->align_buffer = dma_alloc_coherent(mmc_dev(mmc),
> + host->align_buffer_sz,
> + &host->align_addr,
> + GFP_KERNEL);
> host->align_buffer = kmalloc(host->align_buffer_sz, GFP_KERNEL);
kmalloc line is still there
> if (!host->adma_table || !host->align_buffer) {
> if (host->adma_table)
> @@ -3010,7 +2983,11 @@ int sdhci_add_host(struct sdhci_host *host)
> host->adma_table_sz,
> host->adma_table,
> host->adma_addr);
> - kfree(host->align_buffer);
> + if (host->align_buffer)
> + dma_free_coherent(mmc_dev(mmc),
> + host->align_buffer_sz,
> + host->align_buffer,
> + host->align_addr);
> pr_warn("%s: Unable to allocate ADMA buffers - falling back to standard DMA\n",
> mmc_hostname(mmc));
> host->flags &= ~SDHCI_USE_ADMA;
> @@ -3022,10 +2999,14 @@ int sdhci_add_host(struct sdhci_host *host)
> host->flags &= ~SDHCI_USE_ADMA;
> dma_free_coherent(mmc_dev(mmc), host->adma_table_sz,
> host->adma_table, host->adma_addr);
> - kfree(host->align_buffer);
> + dma_free_coherent(mmc_dev(mmc), host->align_buffer_sz,
> + host->align_buffer, host->align_addr);
> host->adma_table = NULL;
> host->align_buffer = NULL;
> }
> +
> + /* dma_alloc_coherent returns page aligned and sized buffers */
> + BUG_ON(host->align_addr & host->align_mask);
It would be nicer not to have any BUG_ON()
> }
>
> /*
> @@ -3489,7 +3470,9 @@ void sdhci_remove_host(struct sdhci_host *host, int dead)
> if (host->adma_table)
> dma_free_coherent(mmc_dev(mmc), host->adma_table_sz,
> host->adma_table, host->adma_addr);
> - kfree(host->align_buffer);
> + if (host->align_buffer)
> + dma_free_coherent(mmc_dev(mmc), host->align_buffer_sz,
> + host->align_buffer, host->align_addr);
>
> host->adma_table = NULL;
> host->align_buffer = NULL;
>
next prev parent reply other threads:[~2015-12-29 13:47 UTC|newest]
Thread overview: 44+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-12-21 11:39 [PATCH v2 00/24] MMC/SDHCI fixes Russell King - ARM Linux
2015-12-21 11:40 ` [PATCH v2 01/24] mmc: core: shut up "voltage-ranges unspecified" pr_info() Russell King
2015-12-21 11:40 ` [PATCH v2 02/24] mmc: core: improve mmc_of_parse_voltage() to return better status Russell King
2015-12-21 11:40 ` [PATCH v2 03/24] mmc: block: shut up "retrying because a re-tune was needed" message Russell King
2015-12-21 11:40 ` [PATCH v2 04/24] mmc: core: report tuning command execution failure reason Russell King
2015-12-21 11:40 ` [PATCH v2 05/24] mmc: sdhci: move initialisation of command error member Russell King
2015-12-21 11:40 ` [PATCH v2 06/24] mmc: sdhci: clean up command error handling Russell King
2015-12-21 11:40 ` [PATCH v2 07/24] mmc: sdhci: command response CRC " Russell King
2015-12-29 13:08 ` Adrian Hunter
2016-01-02 12:25 ` Russell King - ARM Linux
2016-01-04 11:24 ` Adrian Hunter
2016-01-26 13:35 ` Russell King - ARM Linux
2015-12-21 11:41 ` [PATCH v2 08/24] mmc: sdhci: avoid unnecessary mapping/unmapping of align buffer Russell King
2015-12-29 13:44 ` Adrian Hunter [this message]
2016-01-02 12:29 ` Russell King - ARM Linux
2016-01-02 14:31 ` Russell King - ARM Linux
2016-01-04 11:41 ` Adrian Hunter
2016-01-04 11:50 ` Adrian Hunter
2016-01-04 11:56 ` Russell King - ARM Linux
2015-12-21 11:41 ` [PATCH v2 09/24] mmc: sdhci: clean up coding style in sdhci_adma_table_pre() Russell King
2015-12-21 11:41 ` [PATCH v2 10/24] mmc: sdhci: avoid walking SG list for writes Russell King
2015-12-21 11:41 ` [PATCH v2 11/24] mmc: sdhci: factor out common DMA cleanup in sdhci_finish_data() Russell King
2015-12-21 11:41 ` [PATCH v2 12/24] mmc: sdhci: move sdhci_pre_dma_transfer() Russell King
2015-12-21 11:41 ` [PATCH v2 13/24] mmc: sdhci: factor out sdhci_pre_dma_transfer() from sdhci_adma_table_pre() Russell King
2015-12-21 11:41 ` [PATCH v2 14/24] mmc: sdhci: pass the cookie into sdhci_pre_dma_transfer() Russell King
2015-12-21 11:41 ` [PATCH v2 15/24] mmc: sdhci: always unmap a mapped data transfer in sdhci_post_req() Russell King
2015-12-21 11:41 ` [PATCH v2 16/24] mmc: sdhci: clean up host cookie handling Russell King
2015-12-21 11:41 ` [PATCH v2 17/24] mmc: sdhci: plug DMA mapping leak on error Russell King
2015-12-21 11:41 ` [PATCH v2 18/24] mmc: sdhci-pxav3: fix higher speed mode capabilities Russell King
2015-12-21 11:54 ` Marcin Wojtas
2015-12-21 11:41 ` [PATCH v2 19/24] mmc: sdhci: further fix for DMA unmapping in sdhci_post_req() Russell King
2015-12-21 11:42 ` [PATCH v2 20/24] mmc: sdhci: fix data timeout (part 1) Russell King
2015-12-21 11:42 ` [PATCH v2 21/24] mmc: sdhci: fix data timeout (part 2) Russell King
2015-12-21 11:42 ` [PATCH v2 22/24] mmc: sdhci: prepare DMA address/size quirk handling consolidation Russell King
2015-12-21 11:42 ` [PATCH v2 23/24] mmc: sdhci: consolidate the DMA/ADMA size/address quicks Russell King
2015-12-21 11:42 ` [PATCH v2 24/24] mmc: sdhci: further code simplication Russell King
2015-12-21 12:35 ` [PATCH v2 00/24] MMC/SDHCI fixes Ulf Hansson
2015-12-21 12:51 ` Russell King - ARM Linux
2015-12-21 13:23 ` Ulf Hansson
2015-12-21 13:41 ` Russell King - ARM Linux
2015-12-21 13:59 ` Ulf Hansson
2015-12-22 11:25 ` Ulf Hansson
2015-12-22 11:40 ` Russell King - ARM Linux
2015-12-21 12:58 ` Russell King - ARM Linux
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=56828E46.5090600@intel.com \
--to=adrian.hunter@intel.com \
--cc=gregory.clement@free-electrons.com \
--cc=kernel@pengutronix.de \
--cc=linux-mmc@vger.kernel.org \
--cc=mw@semihalf.com \
--cc=rmk+kernel@arm.linux.org.uk \
--cc=shawnguo@kernel.org \
--cc=ulf.hansson@linaro.org \
/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: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).