Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190111110851.6805-1-faiz_abbas@ti.com> <20190111110851.6805-2-faiz_abbas@ti.com>
 <bcea7bae-41c8-358a-11c1-3886f9dc0351@intel.com> <CAG2=9p-HsbghD2-m6VPEPCQVuSWC_1mZ=GL+aKZwZTgOgSO0sQ@mail.gmail.com>
 <a68b68f6-ea5f-2193-188f-cddeec9f6ef4@ti.com> <CAG2=9p_qU8XOtyanEPfa1CSY_X5QGgGCvBg-7dcyW-BhcBChKQ@mail.gmail.com>
 <8e9a58d3-35e0-7b0f-328d-9c049024a6aa@ti.com>
In-Reply-To: <8e9a58d3-35e0-7b0f-328d-9c049024a6aa@ti.com>
From:   Chunyan Zhang <zhang.chunyan@linaro.org>
Date:   Wed, 30 Jan 2019 13:01:46 +0800
Message-ID: <CAG2=9p_mUyY=hKLN4E-ZUaJNueE-h5hHKDnJV5cG5kjtLsEgPA@mail.gmail.com>
Subject: Re: [PATCH 1/7] mmc: sdhci: add support for using external DMA devices
To:     Faiz Abbas <faiz_abbas@ti.com>
Cc:     Adrian Hunter <adrian.hunter@intel.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        devicetree@vger.kernel.org, linux-mmc@vger.kernel.org,
        Kishon Vijay Abraham I <kishon@ti.com>,
        Mark Rutland <mark.rutland@arm.com>,
        Rob Herring <robh+dt@kernel.org>,
        Ulf Hansson <ulf.hansson@linaro.org>,
        Mark Brown <broonie@kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi Faiz=EF=BC=8C

On Tue, 29 Jan 2019 at 19:50, Faiz Abbas <faiz_abbas@ti.com> wrote:
>
> Chunyan,
>
> On 28/01/19 5:16 PM, Chunyan Zhang wrote:
> > On Mon, 28 Jan 2019 at 18:40, Faiz Abbas <faiz_abbas@ti.com> wrote:
> >>
> >> Hi,
> >>
> >> On 28/01/19 3:50 PM, Chunyan Zhang wrote:
> >>> On Thu, 24 Jan 2019 at 19:41, Adrian Hunter <adrian.hunter@intel.com>=
 wrote:
> >>>>
> >>>> On 11/01/19 1:08 PM, Faiz Abbas wrote:
> >>>>> From: Chunyan Zhang <zhang.chunyan@linaro.org>
> >>>>>
> >>>>> Some standard SD host controllers can support both external dma
> >>>>> controllers as well as ADMA/SDMA in which the SD host controller
> >>>>> acts as DMA master. TI's omap controller is the case as an example.
> >>>>>
> >>>>> Currently the generic SDHCI code supports ADMA/SDMA integrated in
> >>>>> the host controller but does not have any support for external DMA
> >>>>> controllers implemented using dmaengine, meaning that custom code i=
s
> >>>>> needed for any systems that use an external DMA controller with SDH=
CI.
> >>>>>
> >>>>> Fixes by Faiz Abbas <faiz_abbas@ti.com>:
> >>>>> 1. Map scatterlists before dmaengine_prep_slave_sg()
> >>>>> 2. Use dma_async() functions inside of the send_command() path and
> >>>>> synchronize once at the start of each request.
> >>>>
> >>>> Sorry for the slow reply, but I do have some concerns.  Please see t=
he comments.
> >>>>
> >>>>>
> >>>>> Signed-off-by: Chunyan Zhang <zhang.chunyan@linaro.org>
> >>>>> Signed-off-by: Faiz Abbas <faiz_abbas@ti.com>
> >>>>> ---
> >>>>>  drivers/mmc/host/Kconfig |   3 +
> >>>>>  drivers/mmc/host/sdhci.c | 266 +++++++++++++++++++++++++++++++++++=
+++-
> >>>>>  drivers/mmc/host/sdhci.h |   8 ++
> >>>>>  3 files changed, 273 insertions(+), 4 deletions(-)
> >>>>>
> >>>>> diff --git a/drivers/mmc/host/Kconfig b/drivers/mmc/host/Kconfig
> >>>>> index e26b8145efb3..333292e8ecdd 100644
> >>>>> --- a/drivers/mmc/host/Kconfig
> >>>>> +++ b/drivers/mmc/host/Kconfig
> >>>>> @@ -999,3 +999,6 @@ config MMC_SDHCI_AM654
> >>>>>         If you have a controller with this interface, say Y or M he=
re.
> >>>>>
> >>>>>         If unsure, say N.
> >>>>> +
> >>>>> +config MMC_SDHCI_EXTERNAL_DMA
> >>>>> +        bool
> >>>>> diff --git a/drivers/mmc/host/sdhci.c b/drivers/mmc/host/sdhci.c
> >>>>> index a22e11a65658..4a9044c06e21 100644
> >>>>> --- a/drivers/mmc/host/sdhci.c
> >>>>> +++ b/drivers/mmc/host/sdhci.c
> >>>>> @@ -14,6 +14,7 @@
> >>>>>   */
> >>>>>
> >>>>>  #include <linux/delay.h>
> >>>>> +#include <linux/dmaengine.h>
> >>>>>  #include <linux/ktime.h>
> >>>>>  #include <linux/highmem.h>
> >>>>>  #include <linux/io.h>
> >>>>> @@ -1118,6 +1119,226 @@ static void sdhci_prepare_data(struct sdhci=
_host *host, struct mmc_command *cmd)
> >>>>>       }
> >>>>>  }
> >>>>>
> >>>>> +#if IS_ENABLED(CONFIG_MMC_SDHCI_EXTERNAL_DMA)
> >>>>> +static int sdhci_external_dma_init(struct sdhci_host *host)
> >>>>> +{
> >>>>> +     int ret =3D 0;
> >>>>> +     struct mmc_host *mmc =3D host->mmc;
> >>>>> +
> >>>>> +     host->tx_chan =3D dma_request_chan(mmc->parent, "tx");
> >>>>> +     if (IS_ERR(host->tx_chan)) {
> >>>>> +             ret =3D PTR_ERR(host->tx_chan);
> >>>>> +             if (ret !=3D -EPROBE_DEFER)
> >>>>> +                     pr_warn("Failed to request TX DMA channel.\n"=
);
> >>>>> +             host->tx_chan =3D NULL;
> >>>>> +             return ret;
> >>>>> +     }
> >>>>> +
> >>>>> +     host->rx_chan =3D dma_request_chan(mmc->parent, "rx");
> >>>>> +     if (IS_ERR(host->rx_chan)) {
> >>>>> +             if (host->tx_chan) {
> >>>>> +                     dma_release_channel(host->tx_chan);
> >>>>> +                     host->tx_chan =3D NULL;
> >>>>> +             }
> >>>>> +
> >>>>> +             ret =3D PTR_ERR(host->rx_chan);
> >>>>> +             if (ret !=3D -EPROBE_DEFER)
> >>>>> +                     pr_warn("Failed to request RX DMA channel.\n"=
);
> >>>>> +             host->rx_chan =3D NULL;
> >>>>> +     }
> >>>>> +
> >>>>> +     return ret;
> >>>>> +}
> >>>>> +
> >>>>> +static inline struct dma_chan *
> >>>>> +sdhci_external_dma_channel(struct sdhci_host *host, struct mmc_dat=
a *data)
> >>>>> +{
> >>>>> +     return data->flags & MMC_DATA_WRITE ? host->tx_chan : host->r=
x_chan;
> >>>>> +}
> >>>>> +
> >>>>> +static int sdhci_external_dma_setup(struct sdhci_host *host,
> >>>>> +                                 struct mmc_command *cmd)
> >>>>> +{
> >>>>> +     int ret, i;
> >>>>> +     struct dma_async_tx_descriptor *desc;
> >>>>> +     struct mmc_data *data =3D cmd->data;
> >>>>> +     struct dma_chan *chan;
> >>>>> +     struct dma_slave_config cfg;
> >>>>> +     dma_cookie_t cookie;
> >>>>> +     int sg_cnt;
> >>>>> +
> >>>>> +     if (!host->mapbase)
> >>>>> +             return -EINVAL;
> >>>>> +
> >>>>> +     cfg.src_addr =3D host->mapbase + SDHCI_BUFFER;
> >>>>> +     cfg.dst_addr =3D host->mapbase + SDHCI_BUFFER;
> >>>>> +     cfg.src_addr_width =3D DMA_SLAVE_BUSWIDTH_4_BYTES;
> >>>>> +     cfg.dst_addr_width =3D DMA_SLAVE_BUSWIDTH_4_BYTES;
> >>>>> +     cfg.src_maxburst =3D data->blksz / 4;
> >>>>> +     cfg.dst_maxburst =3D data->blksz / 4;
> >>>>> +
> >>>>> +     /* Sanity check: all the SG entries must be aligned by block =
size. */
> >>>>> +     for (i =3D 0; i < data->sg_len; i++) {
> >>>>> +             if ((data->sg + i)->length % data->blksz)
> >>>>> +                     return -EINVAL;
> >>>>> +     }
> >>>>> +
> >>>>> +     chan =3D sdhci_external_dma_channel(host, data);
> >>>>> +
> >>>>> +     ret =3D dmaengine_slave_config(chan, &cfg);
> >>>>> +     if (ret)
> >>>>> +             return ret;
> >>>>> +
> >>>>> +     sg_cnt =3D sdhci_pre_dma_transfer(host, data, COOKIE_MAPPED);
> >>>>> +     if (sg_cnt <=3D 0)
> >>>>> +             return -EINVAL;
> >>>>> +
> >>>>> +     desc =3D dmaengine_prep_slave_sg(chan, data->sg, data->sg_len=
,
> >>>>> +                                    mmc_get_dma_dir(data),
> >>>>> +                                    DMA_PREP_INTERRUPT | DMA_CTRL_=
ACK);
> >>>>> +     if (!desc)
> >>>>> +             return -EINVAL;
> >>>>> +
> >>>>> +     desc->callback =3D NULL;
> >>>>> +     desc->callback_param =3D NULL;
> >>>>> +
> >>>>> +     cookie =3D dmaengine_submit(desc);
> >>>>> +     if (cookie < 0)
> >>>>> +             ret =3D cookie;
> >>>>> +
> >>>>> +     return ret;
> >>>>> +}
> >>>>> +
> >>>>> +static void sdhci_external_dma_release(struct sdhci_host *host)
> >>>>> +{
> >>>>> +     if (host->tx_chan) {
> >>>>> +             dma_release_channel(host->tx_chan);
> >>>>> +             host->tx_chan =3D NULL;
> >>>>> +     }
> >>>>> +
> >>>>> +     if (host->rx_chan) {
> >>>>> +             dma_release_channel(host->rx_chan);
> >>>>> +             host->rx_chan =3D NULL;
> >>>>> +     }
> >>>>> +
> >>>>> +     sdhci_switch_external_dma(host, false);
> >>>>> +}
> >>>>> +
> >>>>> +static int __sdhci_external_dma_prepare_data(struct sdhci_host *ho=
st,
> >>>>> +                                          struct mmc_command *cmd)
> >>>>> +{
> >>>>> +     struct mmc_data *data =3D cmd->data;
> >>>>> +
> >>>>> +     host->data_timeout =3D 0;
> >>>>> +
> >>>>> +     if (sdhci_data_line_cmd(cmd))
> >>>>> +             sdhci_set_timeout(host, cmd);
> >>>>> +
> >>>>> +     WARN_ON(host->data);
> >>>>> +
> >>>>> +     /* Sanity checks */
> >>>>> +     WARN_ON(data->blksz * data->blocks > 524288);
> >>>>> +     WARN_ON(data->blksz > host->mmc->max_blk_size);
> >>>>> +     WARN_ON(data->blocks > 65535);
> >>>>> +
> >>>>> +     host->flags |=3D SDHCI_REQ_USE_DMA;
> >>>>> +     host->data =3D data;
> >>>>> +     host->data_early =3D 0;
> >>>>> +     host->data->bytes_xfered =3D 0;
> >>>>> +
> >>>>> +     sdhci_set_transfer_irqs(host);
> >>>>> +
> >>>>> +     /*
> >>>>> +      * For Version 4.10 onwards, if v4 mode is enabled, 32-bit Bl=
ock Count
> >>>>> +      * can be supported, in that case 16-bit block count register=
 must be 0.
> >>>>> +      */
> >>>>> +     if (host->version >=3D SDHCI_SPEC_410 && host->v4_mode &&
> >>>>> +         (host->quirks2 & SDHCI_QUIRK2_USE_32BIT_BLK_CNT)) {
> >>>>> +             if (sdhci_readw(host, SDHCI_BLOCK_COUNT))
> >>>>> +                     sdhci_writew(host, 0, SDHCI_BLOCK_COUNT);
> >>>>> +             sdhci_writew(host, data->blocks, SDHCI_32BIT_BLK_CNT)=
;
> >>>>> +     } else {
> >>>>> +             sdhci_writew(host, data->blocks, SDHCI_BLOCK_COUNT);
> >>>>> +     }
> >>>>
> >>>> It is probably worth factoring out the code that is shared with
> >>>> sdhci_prepare_data() where possible.
> >>>>
> >>>>> +
> >>>>> +     return 0;
> >>>>> +}
> >>>>> +
> >>>>> +static void sdhci_external_dma_prepare_data(struct sdhci_host *hos=
t,
> >>>>> +                                         struct mmc_command *cmd)
> >>>>> +{
> >>>>> +     struct mmc_data *data =3D cmd->data;
> >>>>> +
> >>>>> +     if (!data)
> >>>>> +             return;
> >>>>
> >>>> Even in the !data case, we still need to set up a timeout for comman=
ds with
> >>>> busy waiting.  I suggest checking the !data case before calling
> >>>> sdhci_external_dma_prepare_data()
> >>>
> >>> Ok.
> >>>
> >>>>
> >>>>> +
> >>>>> +     if (sdhci_external_dma_setup(host, cmd) ||
> >>>>> +         __sdhci_external_dma_prepare_data(host, cmd)) {
> >>>>> +             sdhci_external_dma_release(host);
> >>>>> +             pr_err("%s: Cannot use external DMA, switch to the DM=
A/PIO which standard SDHCI provides.\n",
> >>>>> +                    mmc_hostname(host->mmc));
> >>>>> +             sdhci_prepare_data(host, cmd);
> >>>>> +     }
> >>>>> +}
> >>>>> +
> >>>>> +static void sdhci_external_dma_pre_transfer(struct sdhci_host *hos=
t,
> >>>>> +                                         struct mmc_command *cmd)
> >>>>> +{
> >>>>> +     struct dma_chan *chan;
> >>>>> +
> >>>>> +     if (!cmd->data || cmd->opcode =3D=3D MMC_SET_BLOCK_COUNT)
> >>>>
> >>>> MMC_SET_BLOCK_COUNT never has cmd->data and so does not need to be c=
hecked.
> >>>
> >>> Ok.
> >>>
> >>>>
> >>>>> +             return;
> >>>>> +
> >>>>> +     sdhci_writew(host, cmd->data->blksz, SDHCI_BLOCK_SIZE);
> >>>>
> >>>> Block size is set in __sdhci_external_dma_prepare_data() so does it =
need to
> >>>> be set here also.
> >>>
> >>> Ok.
> >>
> >> That is BLOCK_COUNT. This is BLOCK_SIZE.
> >>
> >>>
> >>>>
> >>>>> +     chan =3D sdhci_external_dma_channel(host, cmd->data);
> >>>>> +     if (chan)
> >>>>> +             dma_async_issue_pending(chan);
> >>>>> +}
> >>>>> +
> >>>>> +static int sdhci_external_dma_cleanup(struct sdhci_host *host,
> >>>>> +                                 struct mmc_data *data)
> >>>>
> >>>> Please align parameters with open parenthesis
> >>>>
> >>>>> +{
> >>>>> +     struct dma_chan *chan =3D sdhci_external_dma_channel(host, da=
ta);
> >>>>> +     int ret =3D 0;
> >>>>> +
> >>>>> +     if (chan)
> >>>>> +             ret =3D dmaengine_terminate_async(chan);
> >>>>> +
> >>>>> +     return ret;
> >>>>> +}
> >>>>> +#else
> >>>>> +static int sdhci_external_dma_init(struct sdhci_host *host)
> >>>>> +{
> >>>>> +     return -EOPNOTSUPP;
> >>>>> +}
> >>>>> +
> >>>>> +static void sdhci_external_dma_release(struct sdhci_host *host)
> >>>>> +{}
> >>>>> +
> >>>>> +static void sdhci_external_dma_prepare_data(struct sdhci_host *hos=
t,
> >>>>> +                                         struct mmc_command *cmd)
> >>>>> +{
> >>>>> +     /* If MMC_SDHCI_EXTERNAL_DMA not supported, PIO will be used =
*/
> >>>>> +     sdhci_prepare_data(host, cmd);
> >>>>> +}
> >>>>> +
> >>>>> +static void sdhci_external_dma_pre_transfer(struct sdhci_host *hos=
t,
> >>>>> +                                         struct mmc_command *cmd)
> >>>>> +{}
> >>>>> +
> >>>>> +static int sdhci_external_dma_cleanup(struct sdhci_host *host,
> >>>>> +                                 struct mmc_data *data)
> >>>>
> >>>> Please align parameters with open parenthesis
> >>>>
> >>>>> +{
> >>>>> +     return 0;
> >>>>> +}
> >>>>> +#endif
> >>>>> +
> >>>>> +void sdhci_switch_external_dma(struct sdhci_host *host, bool en)
> >>>>> +{
> >>>>> +     host->use_external_dma =3D en;
> >>>>> +}
> >>>>> +EXPORT_SYMBOL_GPL(sdhci_switch_external_dma);
> >>>>> +
> >>>>>  static inline bool sdhci_auto_cmd12(struct sdhci_host *host,
> >>>>>                                   struct mmc_request *mrq)
> >>>>>  {
> >>>>> @@ -1374,7 +1595,10 @@ void sdhci_send_command(struct sdhci_host *h=
ost, struct mmc_command *cmd)
> >>>>>               host->data_cmd =3D cmd;
> >>>>>       }
> >>>>>
> >>>>> -     sdhci_prepare_data(host, cmd);
> >>>>> +     if (host->use_external_dma)
> >>>>
> >>>> As mentioned above wrt sdhci_external_dma_prepare_data():
> >>>>
> >>>>         if (host->use_external_dma && cmd->data)
> >>>>
> >>>
> >>> Maybe we should move checking the !data case out of both
> >>> _prepare_data, and add it to sdhci_send_command() before calling
> >>> _prepare_data(), that's saying we can do set up a timeout in
> >>> sdhci_send_command().
> >>
> >> Lets just factor out everything in common as Adrian said. The !data,
> >> BLOCK_COUNT writes and busy waiting.
> >
> > Ok, and also BLOCK_SIZE which can be set in _prepare_data().
> >
> >>
> >>>
> >>>>> +             sdhci_external_dma_prepare_data(host, cmd);
> >>>>> +     else
> >>>>> +             sdhci_prepare_data(host, cmd);
> >>>>>
> >>>>>       sdhci_writel(host, cmd->arg, SDHCI_ARGUMENT);
> >>>>>
> >>>>> @@ -1416,6 +1640,9 @@ void sdhci_send_command(struct sdhci_host *ho=
st, struct mmc_command *cmd)
> >>>>>               timeout +=3D 10 * HZ;
> >>>>>       sdhci_mod_timer(host, cmd->mrq, timeout);
> >>>>>
> >>>>> +     if (host->use_external_dma)
> >>>>> +             sdhci_external_dma_pre_transfer(host, cmd);
> >>>>
> >>>> Why is sdhci_external_dma_pre_transfer() needed here - couldn't it b=
e done
> >>>> in sdhci_external_dma_prepare_data()?
> >>>>
> >>>
> >>> I'm not sure dma_async_issue_pending() can be done so early in
> >>> sdhci_external_dma_prepare_data().
> >>>
> >>>>> +
> >>>>>       sdhci_writew(host, SDHCI_MAKE_CMD(cmd->opcode, flags), SDHCI_=
COMMAND);
> >>>>>  }
> >>>>>  EXPORT_SYMBOL_GPL(sdhci_send_command);
> >>>>> @@ -1781,6 +2008,11 @@ void sdhci_request(struct mmc_host *mmc, str=
uct mmc_request *mrq)
> >>>>>
> >>>>>       sdhci_led_activate(host);
> >>>>>
> >>>>> +     if (host->use_external_dma && mrq->data) {
> >>>>> +             struct dma_chan *chan =3D sdhci_external_dma_channel(=
host,
> >>>>> +                                                                mr=
q->data);
> >>>>
> >>>> sdhci_external_dma_channel is not declared if
> >>>> !IS_ENABLED(CONFIG_MMC_SDHCI_EXTERNAL_DMA)
> >>>
> >>> Ok, I guess this can be moved to sdhci_external_dma_prepare_data().
> >>
> >> No. We have to synchronize from a guaranteed non-atomic context. Maybe
> >> add another empty sdhci_external_dma_channel definition in the #else p=
art?
> >>
> >>>
> >>>>
> >>>>> +             dmaengine_synchronize(chan);
> >>>>
> >>>> So this is to cover for using dmaengine_terminate_async()?
> >>
> >> Yes.
> >>
> >>>
> >>> Ok.
> >>>
> >>>>
> >>>>> +     }
> >>>>>       /*
> >>>>>        * Ensure we don't send the STOP for non-SET_BLOCK_COUNTED
> >>>>>        * requests if Auto-CMD12 is enabled.
> >>>>> @@ -2658,6 +2890,8 @@ static bool sdhci_request_done(struct sdhci_h=
ost *host)
> >>>>>                               dma_unmap_sg(mmc_dev(host->mmc), data=
->sg,
> >>>>>                                            data->sg_len,
> >>>>>                                            mmc_get_dma_dir(data));
> >>>>> +                             if (host->use_external_dma)
> >>>>> +                                     sdhci_external_dma_cleanup(ho=
st, data);
> >>>>
> >>>> Is sdhci_external_dma_cleanup() only needed in the error case?
> >>>>
> >>>> The DMA must be stopped before the memory is unmapped and potentiall=
y freed.
> >>>>
> >>>> Isn't the DMA cleanup also needed in the bounce buffer case?
> >>>>
> >>>> Isn't the DMA cleanup also needed in the COOKIE_PRE_MAPPED case?
> >>>>
> >>>> dmaengine_terminate_async() doesn't stop the DMA but
> >>>> dmaengine_terminate_sync() is not atomic, which looks like a problem=
.
> >>>>
> >>>> Perhaps you look at scheduling some work for the external dma error =
case
> >>>> instead of calling __sdhci_finish_mrq()?  Then the work can do the
> >>>> dmaengine_terminate_sync() and call __sdhci_finish_mrq().
> >>>
> >>> Ok, I will look at these issues.
> >>>
> >>>>
> >>>>>                       }
> >>>>>                       data->host_cookie =3D COOKIE_UNMAPPED;
> >>>>>               }
> >>>>> @@ -3692,12 +3926,15 @@ int sdhci_setup_host(struct sdhci_host *hos=
t)
> >>>>>                      mmc_hostname(mmc), host->version);
> >>>>>       }
> >>>>>
> >>>>> -     if (host->quirks & SDHCI_QUIRK_FORCE_DMA)
> >>>>> +     if (host->quirks & SDHCI_QUIRK_FORCE_DMA) {
> >>>>>               host->flags |=3D SDHCI_USE_SDMA;
> >>>>> -     else if (!(host->caps & SDHCI_CAN_DO_SDMA))
> >>>>> +     } else if (!(host->caps & SDHCI_CAN_DO_SDMA)) {
> >>>>>               DBG("Controller doesn't have SDMA capability\n");
> >>>>> -     else
> >>>>> +     } else if (host->use_external_dma) {
> >>>>> +             /* Using dma-names to detect external dma capability =
*/
> >>>>> +     } else {
> >>>>>               host->flags |=3D SDHCI_USE_SDMA;
> >>>>> +     }
> >>>>
> >>>> These if-statements are about setting SDHCI_USE_SDMA but why is a ch=
ange
> >>>> needed for the host->use_external_dma case?
> >>>
> >>> Yes, this is not needed, otherwise the controller cannot switch back
> >>> to SDMA if it is supported in the controller.
> >>
> >> The else case means that the host always uses SDMA. Needed to make sur=
e
> >> this doesn't happen in the external_dma case. Its not SDHCI_CAN_DO_SDM=
A,
> >> its SDHCI_USE_SDMA.
> >>
> >>>
> >>>>
> >>>>>
> >>>>>       if ((host->quirks & SDHCI_QUIRK_BROKEN_DMA) &&
> >>>>>               (host->flags & SDHCI_USE_SDMA)) {
> >>>>> @@ -3785,6 +4022,19 @@ int sdhci_setup_host(struct sdhci_host *host=
)
> >>>>>               }
> >>>>>       }
> >>>>>
> >>>>> +     if (host->use_external_dma) {
> >>>>> +             ret =3D sdhci_external_dma_init(host);
> >>>>> +             if (ret =3D=3D -EPROBE_DEFER)
> >>>>> +                     goto unreg;
> >>>>> +
> >>>>> +             /*
> >>>>> +              * Fall back to use the DMA/PIO integrated in standar=
d SDHCI
> >>>>> +              * instead of external DMA devices.
> >>>>> +              */
> >>>>> +             if (ret)
> >>>>> +                     sdhci_switch_external_dma(host, false);
> >>>>> +     }
> >>>>> +
> >>>>>       /*
> >>>>>        * If we use DMA, then it's up to the caller to set the DMA
> >>>>>        * mask, but PIO does not need the hw shim so we set a new
> >>>>> @@ -4201,6 +4451,10 @@ void sdhci_cleanup_host(struct sdhci_host *h=
ost)
> >>>>>               dma_free_coherent(mmc_dev(mmc), host->align_buffer_sz=
 +
> >>>>>                                 host->adma_table_sz, host->align_bu=
ffer,
> >>>>>                                 host->align_addr);
> >>>>> +
> >>>>> +     if (host->use_external_dma)
> >>>>> +             sdhci_external_dma_release(host);
> >>>>> +
> >>>>>       host->adma_table =3D NULL;
> >>>>>       host->align_buffer =3D NULL;
> >>>>>  }
> >>>>> @@ -4247,6 +4501,7 @@ int __sdhci_add_host(struct sdhci_host *host)
> >>>>>
> >>>>>       pr_info("%s: SDHCI controller on %s [%s] using %s\n",
> >>>>>               mmc_hostname(mmc), host->hw_name, dev_name(mmc_dev(mm=
c)),
> >>>>> +             host->use_external_dma ? "External DMA" :
> >>>>>               (host->flags & SDHCI_USE_ADMA) ?
> >>>>>               (host->flags & SDHCI_USE_64_BIT_DMA) ? "ADMA 64-bit" =
: "ADMA" :
> >>>>>               (host->flags & SDHCI_USE_SDMA) ? "DMA" : "PIO");
> >>>>> @@ -4335,6 +4590,9 @@ void sdhci_remove_host(struct sdhci_host *hos=
t, int dead)
> >>>>>                                 host->adma_table_sz, host->align_bu=
ffer,
> >>>>>                                 host->align_addr);
> >>>>>
> >>>>> +     if (host->use_external_dma)
> >>>>> +             sdhci_external_dma_release(host);
> >>>>> +
> >>>>>       host->adma_table =3D NULL;
> >>>>>       host->align_buffer =3D NULL;
> >>>>>  }
> >>>>> diff --git a/drivers/mmc/host/sdhci.h b/drivers/mmc/host/sdhci.h
> >>>>> index 6cc9a3c2ac66..7a52823ebef4 100644
> >>>>> --- a/drivers/mmc/host/sdhci.h
> >>>>> +++ b/drivers/mmc/host/sdhci.h
> >>>>> @@ -482,6 +482,7 @@ struct sdhci_host {
> >>>>>
> >>>>>       int irq;                /* Device IRQ */
> >>>>>       void __iomem *ioaddr;   /* Mapped address */
> >>>>> +     phys_addr_t mapbase;    /* physical address base */
> >>>>>       char *bounce_buffer;    /* For packing SDMA reads/writes */
> >>>>>       dma_addr_t bounce_addr;
> >>>>>       unsigned int bounce_buffer_size;
> >>>>> @@ -531,6 +532,7 @@ struct sdhci_host {
> >>>>>       bool pending_reset;     /* Cmd/data reset is pending */
> >>>>>       bool irq_wake_enabled;  /* IRQ wakeup is enabled */
> >>>>>       bool v4_mode;           /* Host Version 4 Enable */
> >>>>> +     bool use_external_dma; /* Host selects to use external DMA */
> >>>>
> >>>> Please align /**/ with above i.e. use tab
> >>
> >> Will fix.
> >
> > Do you need me to address these comments and then send you a patch or
> > you'd like to do instead?
> >
>
> This would be much appreciated. Would you be able to work on it this
> week? I would like to send a v2 by Friday.

I will try. I just have worries about handling the external dma error case.
Anyway I'll send you a patch by tomorrow, since I will leave 10 days
from this Friday.

Thanks,
Chunyan