This reverts commit 5dd195522562542bc6ebe6e7bd47890d8b7ca93c.
First, the fix seems to be plain wrong, since the erratum suggests
waiting 5ms before setting setting SYSCTL[RSTD], but this msleep()
happens after the call of sdhci_reset() which is where that bit gets
set (if SDHCI_RESET_DATA is in mask).
Second, walking the whole device tree to figure out if some node has a
"fsl,p2020-esdhc" compatible string is hugely expensive - about 70 to
100 us on our mpc8309 board. Walking the device tree is done under a
raw_spin_lock, so this is obviously really bad on an -rt system, and a
waste of time on all.
In fact, since esdhc_reset() seems to get called around 100 times per
second, that mpc8309 now spends 0.8% of its time determining that
it is not a p2020. Whether those 100 calls/s are normal or due to some
other bug or misconfiguration, regularly hitting a 100 us
non-preemptible window is unacceptable.
Signed-off-by: Rasmus Villemoes <[email protected]>
---
The errata sheet for mpc8309 also mentions A-009204, so I'm not at all
opposed to having a fix for that. But it needs to be done properly
without causing a huge performance or latency impact. We should
probably just add a bit to struct sdhci_esdhc which gets initialized
in esdhc_init.
drivers/mmc/host/sdhci-of-esdhc.c | 3 ---
1 file changed, 3 deletions(-)
diff --git a/drivers/mmc/host/sdhci-of-esdhc.c b/drivers/mmc/host/sdhci-of-esdhc.c
index 5cca3fa4610b..7f87a90bf56a 100644
--- a/drivers/mmc/host/sdhci-of-esdhc.c
+++ b/drivers/mmc/host/sdhci-of-esdhc.c
@@ -764,9 +764,6 @@ static void esdhc_reset(struct sdhci_host *host, u8 mask)
sdhci_writel(host, host->ier, SDHCI_INT_ENABLE);
sdhci_writel(host, host->ier, SDHCI_SIGNAL_ENABLE);
- if (of_find_compatible_node(NULL, NULL, "fsl,p2020-esdhc"))
- mdelay(5);
-
if (mask & SDHCI_RESET_ALL) {
val = sdhci_readl(host, ESDHC_TBCTL);
val &= ~ESDHC_TB_EN;
--
2.23.0
ping, any comments on this revert?
On 04/12/2019 09.54, Rasmus Villemoes wrote:
> This reverts commit 5dd195522562542bc6ebe6e7bd47890d8b7ca93c.
>
> First, the fix seems to be plain wrong, since the erratum suggests
> waiting 5ms before setting setting SYSCTL[RSTD], but this msleep()
> happens after the call of sdhci_reset() which is where that bit gets
> set (if SDHCI_RESET_DATA is in mask).
On Wed, 4 Dec 2019 at 09:54, Rasmus Villemoes <[email protected]> wrote:
>
> This reverts commit 5dd195522562542bc6ebe6e7bd47890d8b7ca93c.
>
> First, the fix seems to be plain wrong, since the erratum suggests
> waiting 5ms before setting setting SYSCTL[RSTD], but this msleep()
> happens after the call of sdhci_reset() which is where that bit gets
> set (if SDHCI_RESET_DATA is in mask).
>
> Second, walking the whole device tree to figure out if some node has a
> "fsl,p2020-esdhc" compatible string is hugely expensive - about 70 to
> 100 us on our mpc8309 board. Walking the device tree is done under a
> raw_spin_lock, so this is obviously really bad on an -rt system, and a
> waste of time on all.
>
> In fact, since esdhc_reset() seems to get called around 100 times per
> second, that mpc8309 now spends 0.8% of its time determining that
> it is not a p2020. Whether those 100 calls/s are normal or due to some
> other bug or misconfiguration, regularly hitting a 100 us
> non-preemptible window is unacceptable.
>
> Signed-off-by: Rasmus Villemoes <[email protected]>
Applied for fixes and by adding a stable tag, thanks!
Kind regards
Uffe
> ---
>
> The errata sheet for mpc8309 also mentions A-009204, so I'm not at all
> opposed to having a fix for that. But it needs to be done properly
> without causing a huge performance or latency impact. We should
> probably just add a bit to struct sdhci_esdhc which gets initialized
> in esdhc_init.
>
> drivers/mmc/host/sdhci-of-esdhc.c | 3 ---
> 1 file changed, 3 deletions(-)
>
> diff --git a/drivers/mmc/host/sdhci-of-esdhc.c b/drivers/mmc/host/sdhci-of-esdhc.c
> index 5cca3fa4610b..7f87a90bf56a 100644
> --- a/drivers/mmc/host/sdhci-of-esdhc.c
> +++ b/drivers/mmc/host/sdhci-of-esdhc.c
> @@ -764,9 +764,6 @@ static void esdhc_reset(struct sdhci_host *host, u8 mask)
> sdhci_writel(host, host->ier, SDHCI_INT_ENABLE);
> sdhci_writel(host, host->ier, SDHCI_SIGNAL_ENABLE);
>
> - if (of_find_compatible_node(NULL, NULL, "fsl,p2020-esdhc"))
> - mdelay(5);
> -
> if (mask & SDHCI_RESET_ALL) {
> val = sdhci_readl(host, ESDHC_TBCTL);
> val &= ~ESDHC_TB_EN;
> --
> 2.23.0
>
> -----Original Message-----
> From: [email protected] <[email protected]>
> On Behalf Of Ulf Hansson
> Sent: Wednesday, December 18, 2019 9:40 PM
> To: Yinbo Zhu <[email protected]>; Y.b. Lu <[email protected]>
> Cc: Adrian Hunter <[email protected]>; Rasmus Villemoes
> <[email protected]>; [email protected]; Linux Kernel
> Mailing List <[email protected]>
> Subject: Re: [PATCH] mmc: sdhci-of-esdhc: Revert "mmc: sdhci-of-esdhc: add
> erratum A-009204 support"
>
> Yinbo, Yangbo
>
> On Tue, 10 Dec 2019 at 10:51, Ulf Hansson <[email protected]> wrote:
> >
> > On Wed, 4 Dec 2019 at 09:54, Rasmus Villemoes
> <[email protected]> wrote:
> > >
> > > This reverts commit 5dd195522562542bc6ebe6e7bd47890d8b7ca93c.
> > >
> > > First, the fix seems to be plain wrong, since the erratum suggests
> > > waiting 5ms before setting setting SYSCTL[RSTD], but this msleep()
> > > happens after the call of sdhci_reset() which is where that bit gets
> > > set (if SDHCI_RESET_DATA is in mask).
> > >
> > > Second, walking the whole device tree to figure out if some node has a
> > > "fsl,p2020-esdhc" compatible string is hugely expensive - about 70 to
> > > 100 us on our mpc8309 board. Walking the device tree is done under a
> > > raw_spin_lock, so this is obviously really bad on an -rt system, and a
> > > waste of time on all.
> > >
> > > In fact, since esdhc_reset() seems to get called around 100 times per
> > > second, that mpc8309 now spends 0.8% of its time determining that
> > > it is not a p2020. Whether those 100 calls/s are normal or due to some
> > > other bug or misconfiguration, regularly hitting a 100 us
> > > non-preemptible window is unacceptable.
> > >
> > > Signed-off-by: Rasmus Villemoes <[email protected]>
> >
> > Applied for fixes and by adding a stable tag, thanks!
>
> Just wanted to highlight, that $subject patch has been applied for
> fixes, which means we need a new fix the errata A-009204.
>
> Rasmus kind of already hinted on how this could be fixed in a better
> way, hope this helps.
>
Sorry Uffe, I missed the patch since my name wasn't in TO/CC list.
I have sent out a patch for re-implementation.
Thanks!
> Kind regards
> Uffe
>
> > > ---
> > >
> > > The errata sheet for mpc8309 also mentions A-009204, so I'm not at all
> > > opposed to having a fix for that. But it needs to be done properly
> > > without causing a huge performance or latency impact. We should
> > > probably just add a bit to struct sdhci_esdhc which gets initialized
> > > in esdhc_init.
> > >
> > > drivers/mmc/host/sdhci-of-esdhc.c | 3 ---
> > > 1 file changed, 3 deletions(-)
> > >
> > > diff --git a/drivers/mmc/host/sdhci-of-esdhc.c
> b/drivers/mmc/host/sdhci-of-esdhc.c
> > > index 5cca3fa4610b..7f87a90bf56a 100644
> > > --- a/drivers/mmc/host/sdhci-of-esdhc.c
> > > +++ b/drivers/mmc/host/sdhci-of-esdhc.c
> > > @@ -764,9 +764,6 @@ static void esdhc_reset(struct sdhci_host *host, u8
> mask)
> > > sdhci_writel(host, host->ier, SDHCI_INT_ENABLE);
> > > sdhci_writel(host, host->ier, SDHCI_SIGNAL_ENABLE);
> > >
> > > - if (of_find_compatible_node(NULL, NULL, "fsl,p2020-esdhc"))
> > > - mdelay(5);
> > > -
> > > if (mask & SDHCI_RESET_ALL) {
> > > val = sdhci_readl(host, ESDHC_TBCTL);
> > > val &= ~ESDHC_TB_EN;
> > > --
> > > 2.23.0
> > >