| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <15c31e3b-646a-f5db-b45a-3e2f556d9fd5@st.com>
Date: Mon, 24 Sep 2018 18:12:44 +0200
From: Ludovic BARRE <ludovic.barre@...com>
To: Ulf Hansson <ulf.hansson@...aro.org>,
Srinivas Kandagatla <srinivas.kandagatla@...aro.org>
CC: Rob Herring <robh+dt@...nel.org>,
Maxime Coquelin <mcoquelin.stm32@...il.com>,
Alexandre Torgue <alexandre.torgue@...com>,
Benjamin Gaignard <benjamin.gaignard@...aro.org>,
Gerald Baeza <gerald.baeza@...com>,
Loic Pallardy <loic.pallardy@...com>,
Linux ARM <linux-arm-kernel@...ts.infradead.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
DTML <devicetree@...r.kernel.org>,
"linux-mmc@...r.kernel.org" <linux-mmc@...r.kernel.org>,
<linux-stm32@...md-mailman.stormreply.com>
Subject: Re: [PATCH V2 03/27] mmc: mmci: convert dma_setup callback to return
an int
hi Ulf
On 09/24/2018 04:28 PM, Ulf Hansson wrote:
> On 21 September 2018 at 11:45, Ludovic Barre <ludovic.Barre@...com> wrote:
>> From: Ludovic Barre <ludovic.barre@...com>
>>
>> This patch converts dma_setup callback to return an integer
>> This patch is needed to prepare sdmmc variant with internal dma
>>
>> Signed-off-by: Ludovic Barre <ludovic.barre@...com>
>> ---
>> drivers/mmc/host/mmci.c | 14 ++++++++++----
>> drivers/mmc/host/mmci.h | 2 +-
>> drivers/mmc/host/mmci_qcom_dml.c | 6 ++++--
>> 3 files changed, 15 insertions(+), 7 deletions(-)
>>
>> diff --git a/drivers/mmc/host/mmci.c b/drivers/mmc/host/mmci.c
>> index cbd67bc..2f845f3 100644
>> --- a/drivers/mmc/host/mmci.c
>> +++ b/drivers/mmc/host/mmci.c
>> @@ -415,7 +415,7 @@ static void mmci_init_sg(struct mmci_host *host, struct mmc_data *data)
>> * no custom DMA interfaces are supported.
>> */
>> #ifdef CONFIG_DMA_ENGINE
>> -static void mmci_dma_setup(struct mmci_host *host)
>> +static int mmci_dma_setup(struct mmci_host *host)
>> {
>> const char *rxname, *txname;
>>
>> @@ -466,7 +466,9 @@ static void mmci_dma_setup(struct mmci_host *host)
>> }
>>
>> if (host->ops && host->ops->dma_setup)
>> - host->ops->dma_setup(host);
>> + return host->ops->dma_setup(host);
>> +
>
> Looks like you need to implement a error handling, instead of just
> returning the error code.
yes, today if a dma_setup is defined, we can't fallback to pio mode.
A use_dma variable (defined in mmci_host) could be setted follow the
return code of ops->dma_setup
example:
void mmci_dma_setup(struct mmci_host *host)
{
int err;
if (!host->ops || !host->ops->dma_setup)
return;
err = host->ops->dma_setup(host);
if (err)
return;
host->use_dma = true;
/* initialize pre request cookie */
host->next_cookie = 1;
}
use_dma variable could be check by mmci_host_ops variant functions.
are you agree with this proposal?
>
>> + return 0;
>> }
>>
>> /*
>> @@ -741,8 +743,10 @@ static void mmci_post_request(struct mmc_host *mmc, struct mmc_request *mrq,
>> static void mmci_get_next_data(struct mmci_host *host, struct mmc_data *data)
>> {
>> }
>> -static inline void mmci_dma_setup(struct mmci_host *host)
>> +
>> +static inline int mmci_dma_setup(struct mmci_host *host)
>> {
>> + return 0;
>> }
>>
>> static inline void mmci_dma_release(struct mmci_host *host)
>> @@ -1796,7 +1800,9 @@ static int mmci_probe(struct amba_device *dev,
>> amba_rev(dev), (unsigned long long)dev->res.start,
>> dev->irq[0], dev->irq[1]);
>>
>> - mmci_dma_setup(host);
>> + ret = mmci_dma_setup(host);
>> + if (ret)
>> + goto clk_disable;
>>
>> pm_runtime_set_autosuspend_delay(&dev->dev, 50);
>> pm_runtime_use_autosuspend(&dev->dev);
>> diff --git a/drivers/mmc/host/mmci.h b/drivers/mmc/host/mmci.h
>> index 21aaf9a..06299ac 100644
>> --- a/drivers/mmc/host/mmci.h
>> +++ b/drivers/mmc/host/mmci.h
>> @@ -273,7 +273,7 @@ struct variant_data {
>>
>> /* mmci variant callbacks */
>> struct mmci_host_ops {
>> - void (*dma_setup)(struct mmci_host *host);
>> + int (*dma_setup)(struct mmci_host *host);
>> };
>>
>> struct mmci_host_next {
>> diff --git a/drivers/mmc/host/mmci_qcom_dml.c b/drivers/mmc/host/mmci_qcom_dml.c
>> index be3fab5..1bb59cf 100644
>> --- a/drivers/mmc/host/mmci_qcom_dml.c
>> +++ b/drivers/mmc/host/mmci_qcom_dml.c
>> @@ -119,7 +119,7 @@ static int of_get_dml_pipe_index(struct device_node *np, const char *name)
>> }
>>
>> /* Initialize the dml hardware connected to SD Card controller */
>> -static void qcom_dma_setup(struct mmci_host *host)
>> +static int qcom_dma_setup(struct mmci_host *host)
>> {
>> u32 config;
>> void __iomem *base;
>> @@ -131,7 +131,7 @@ static void qcom_dma_setup(struct mmci_host *host)
>>
>> if (producer_id < 0 || consumer_id < 0) {
>> host->variant->qcom_dml = false;
>
> It seems like the existing code is an attempt to fallback to use pio
> mode. However, I doubt it works as is.
oops, I seen that like a variant identification issue, so if it's a
fallback to pio mode it doesn't work.
>
>> - return;
>> + return -EINVAL;
>
> My point is, if you return an error code here, it means that the error
> code becomes propagated and ->probe() will fail.
>
> Ideally, we should be able fall back to pio mode when dma doesn't
> work. I have looped in Srinivas who implemented the qcom dml support,
> let's see if he can explains the intent with the code.
>
> I also volunteer to help out running some tests on the 410c platform,
> however allow me a day or two to do that.
yes, I would be more reassured
>
>> }
>>
>> base = host->base + DML_OFFSET;
>> @@ -175,6 +175,8 @@ static void qcom_dma_setup(struct mmci_host *host)
>>
>> /* Make sure dml initialization is finished */
>> mb();
>> +
>> + return 0;
>> }
>>
>> static struct mmci_host_ops qcom_variant_ops = {
>> --
>> 2.7.4
>>
>
> Kind regards
> Uffe
>
Powered by blists - more mailing lists