linux-kernel - Re: [PATCH 1/4] mmc: mmci: avoid fake busy polling

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
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Thu, 25 Apr 2019 11:22:37 +0200
From:   Ludovic BARRE <ludovic.barre@...com>
To:     Ulf Hansson <ulf.hansson@...aro.org>
CC:     Rob Herring <robh+dt@...nel.org>,
        Srinivas Kandagatla <srinivas.kandagatla@...aro.org>,
        Maxime Coquelin <mcoquelin.stm32@...il.com>,
        Alexandre Torgue <alexandre.torgue@...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 1/4] mmc: mmci: avoid fake busy polling

hi Ulf

On 4/23/19 3:39 PM, Ulf Hansson wrote:
> On Tue, 5 Mar 2019 at 17:10, Ludovic Barre <ludovic.Barre@...com> wrote:
>>
>> From: Ludovic Barre <ludovic.barre@...com>
>>
>> The busy status bit could occurred even if no busy response is
>> expected (example cmd11). On sdmmc variant, the busy_detect_flag
>> reflects inverted value of d0 state, it's sampled at the end of a
>> CMD response and a second time 2 clk cycles after the CMD response.
>> To avoid a fake busy, the busy status could be checked and polled
>> only if the command has RSP_BUSY flag.
> 
> I would appreciate a better explanation of what this patch really changes.
> 
> The above is giving some background to the behavior of sdmmc variant,
> but at this point that variant doesn't even have the
> ->variant->busy_detect flag set.
> 

Yes, I will try to explain more and focus on common behavior.

>>
>> Signed-off-by: Ludovic Barre <ludovic.barre@...com>
>> ---
>>   drivers/mmc/host/mmci.c | 19 +++++++++++++------
>>   1 file changed, 13 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/mmc/host/mmci.c b/drivers/mmc/host/mmci.c
>> index 387ff14..4901b73 100644
>> --- a/drivers/mmc/host/mmci.c
>> +++ b/drivers/mmc/host/mmci.c
>> @@ -1220,12 +1220,13 @@ mmci_cmd_irq(struct mmci_host *host, struct mmc_command *cmd,
>>               unsigned int status)
>>   {
>>          void __iomem *base = host->base;
>> -       bool sbc;
>> +       bool sbc, busy_resp;
>>
>>          if (!cmd)
>>                  return;
>>
>>          sbc = (cmd == host->mrq->sbc);
>> +       busy_resp = !!(cmd->flags & MMC_RSP_BUSY);
>>
>>          /*
>>           * We need to be one of these interrupts to be considered worth
>> @@ -1239,8 +1240,7 @@ mmci_cmd_irq(struct mmci_host *host, struct mmc_command *cmd,
>>          /*
>>           * ST Micro variant: handle busy detection.
>>           */
>> -       if (host->variant->busy_detect) {
>> -               bool busy_resp = !!(cmd->flags & MMC_RSP_BUSY);
>> +       if (busy_resp && host->variant->busy_detect) {
>>
>>                  /* We are busy with a command, return */
>>                  if (host->busy_status &&
>> @@ -1253,7 +1253,7 @@ mmci_cmd_irq(struct mmci_host *host, struct mmc_command *cmd,
>>                   * that the special busy status bit is still set before
>>                   * proceeding.
>>                   */
>> -               if (!host->busy_status && busy_resp &&
>> +               if (!host->busy_status &&
>>                      !(status & (MCI_CMDCRCFAIL|MCI_CMDTIMEOUT)) &&
>>                      (readl(base + MMCISTATUS) & host->variant->busy_detect_flag)) {
> 
> All the changes above makes perfect sense to me, but looks more like a
> cleanup of the code, rather than actually changing the behavior.

yes, few changing (this just avoid to enter in
"if (host->variant->busy_detect)") at each time.
I could move this part in cleanup patch (before this patch)

> 
>>
>> @@ -1508,6 +1508,7 @@ static irqreturn_t mmci_irq(int irq, void *dev_id)
>>   {
>>          struct mmci_host *host = dev_id;
>>          u32 status;
>> +       bool busy_resp;
>>          int ret = 0;
>>
>>          spin_lock(&host->lock);
>> @@ -1550,9 +1551,15 @@ static irqreturn_t mmci_irq(int irq, void *dev_id)
>>                  }
>>
>>                  /*
>> -                * Don't poll for busy completion in irq context.
>> +                * Don't poll for:
>> +                * -busy completion in irq context.
>> +                * -no busy response expected.
>>                   */
>> -               if (host->variant->busy_detect && host->busy_status)
>> +               busy_resp = host->cmd ?
>> +                       !!(host->cmd->flags & MMC_RSP_BUSY) : false;
> 
> This doesn't make sense to me, but I may be missing something.
> 
> host->busy_status is being updated by mmci_cmd_irq() and only when
> MMC_RSP_BUSY is set for the command in flight. In other words,
> checking for MMC_RSP_BUSY here as well is redundant. No?

In mmci_irq the "do while" loops until the status is totally cleared.

Today (for variant with busy_detect option), the status busy_detect_flag
is excluded only while busy_status period (command with MMC_RSP_BUSY and 
while busy line is low => "busy_status=1")

On SDMMC variant I noticed that busy_detect_flag status could be enabled
even if the command is not MMC_RSP_BUSY, for example sdmmc variant stay
in loop while cmd11 voltage switch.

So I wish extend host->variant->busy_detect_flag exclusion for all
commands which is not a MMC_RSP_BUSY. I suppose that other variants
could have the same behavior, and else there is no impact, normally.

> 
>> +
>> +               if (host->variant->busy_detect &&
>> +                   (!busy_resp || host->busy_status))
>>                          status &= ~host->variant->busy_detect_flag;
>>
>>                  ret = 1;
>> --
>> 2.7.4
>>
> 
> Kind regards
> Uffe
>