Two SBE FIFO fixes by Joachim.
Joachim Fenkes (2):
fsi/sbefifo: Clean up correct FIFO when receiving reset request from
SBE
fsi/sbefifo: Fix reset timeout
drivers/fsi/fsi-sbefifo.c | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
--
2.27.0
From: Joachim Fenkes <[email protected]>
On BMCs with lower timer resolution than 1ms, msleep(1) will take
way longer than 1ms, so looping 10k times won't wait for 10s but
significantly longer.
Fix this by using jiffies like the rest of the code.
Fixes: 9f4a8a2d7f9d ("fsi/sbefifo: Add driver for the SBE FIFO")
Signed-off-by: Joachim Fenkes <[email protected]>
Signed-off-by: Joel Stanley <[email protected]>
---
drivers/fsi/fsi-sbefifo.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/drivers/fsi/fsi-sbefifo.c b/drivers/fsi/fsi-sbefifo.c
index 655b45c1f6ba..3ad9510ad4a4 100644
--- a/drivers/fsi/fsi-sbefifo.c
+++ b/drivers/fsi/fsi-sbefifo.c
@@ -325,6 +325,7 @@ static int sbefifo_up_write(struct sbefifo *sbefifo, __be32 word)
static int sbefifo_request_reset(struct sbefifo *sbefifo)
{
struct device *dev = &sbefifo->fsi_dev->dev;
+ unsigned long end_time;
u32 status, timeout;
int rc;
@@ -341,7 +342,8 @@ static int sbefifo_request_reset(struct sbefifo *sbefifo)
}
/* Wait for it to complete */
- for (timeout = 0; timeout < SBEFIFO_RESET_TIMEOUT; timeout++) {
+ end_time = jiffies + msecs_to_jiffies(SBEFIFO_RESET_TIMEOUT);
+ while (!time_after(jiffies, end_time)) {
rc = sbefifo_regr(sbefifo, SBEFIFO_UP | SBEFIFO_STS, &status);
if (rc) {
dev_err(dev, "Failed to read UP fifo status during reset"
@@ -355,7 +357,7 @@ static int sbefifo_request_reset(struct sbefifo *sbefifo)
return 0;
}
- msleep(1);
+ cond_resched();
}
dev_err(dev, "FIFO reset timed out\n");
--
2.27.0
From: Joachim Fenkes <[email protected]>
When the SBE requests a reset via the down FIFO, that is also the
FIFO we should go and reset ;)
Fixes: 9f4a8a2d7f9d ("fsi/sbefifo: Add driver for the SBE FIFO")
Signed-off-by: Joachim Fenkes <[email protected]>
Signed-off-by: Joel Stanley <[email protected]>
---
drivers/fsi/fsi-sbefifo.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/fsi/fsi-sbefifo.c b/drivers/fsi/fsi-sbefifo.c
index f54df9ebc8b3..655b45c1f6ba 100644
--- a/drivers/fsi/fsi-sbefifo.c
+++ b/drivers/fsi/fsi-sbefifo.c
@@ -400,7 +400,7 @@ static int sbefifo_cleanup_hw(struct sbefifo *sbefifo)
/* The FIFO already contains a reset request from the SBE ? */
if (down_status & SBEFIFO_STS_RESET_REQ) {
dev_info(dev, "Cleanup: FIFO reset request set, resetting\n");
- rc = sbefifo_regw(sbefifo, SBEFIFO_UP, SBEFIFO_PERFORM_RESET);
+ rc = sbefifo_regw(sbefifo, SBEFIFO_DOWN, SBEFIFO_PERFORM_RESET);
if (rc) {
sbefifo->broken = true;
dev_err(dev, "Cleanup: Reset reg write failed, rc=%d\n", rc);
--
2.27.0
On 7/24/20 12:15 AM, Joel Stanley wrote:
> From: Joachim Fenkes <[email protected]>
>
> On BMCs with lower timer resolution than 1ms, msleep(1) will take
> way longer than 1ms, so looping 10k times won't wait for 10s but
> significantly longer.
>
> Fix this by using jiffies like the rest of the code.
>
> Fixes: 9f4a8a2d7f9d ("fsi/sbefifo: Add driver for the SBE FIFO")
> Signed-off-by: Joachim Fenkes <[email protected]>
> Signed-off-by: Joel Stanley <[email protected]>
> ---
> drivers/fsi/fsi-sbefifo.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/fsi/fsi-sbefifo.c b/drivers/fsi/fsi-sbefifo.c
> index 655b45c1f6ba..3ad9510ad4a4 100644
> --- a/drivers/fsi/fsi-sbefifo.c
> +++ b/drivers/fsi/fsi-sbefifo.c
> @@ -325,6 +325,7 @@ static int sbefifo_up_write(struct sbefifo *sbefifo, __be32 word)
> static int sbefifo_request_reset(struct sbefifo *sbefifo)
> {
> struct device *dev = &sbefifo->fsi_dev->dev;
> + unsigned long end_time;
> u32 status, timeout;
> int rc;
>
> @@ -341,7 +342,8 @@ static int sbefifo_request_reset(struct sbefifo *sbefifo)
> }
>
> /* Wait for it to complete */
> - for (timeout = 0; timeout < SBEFIFO_RESET_TIMEOUT; timeout++) {
> + end_time = jiffies + msecs_to_jiffies(SBEFIFO_RESET_TIMEOUT);
> + while (!time_after(jiffies, end_time)) {
> rc = sbefifo_regr(sbefifo, SBEFIFO_UP | SBEFIFO_STS, &status);
> if (rc) {
> dev_err(dev, "Failed to read UP fifo status during reset"
> @@ -355,7 +357,7 @@ static int sbefifo_request_reset(struct sbefifo *sbefifo)
> return 0;
> }
>
> - msleep(1);
> + cond_resched();
A hot loop ? Are you sure ? usleep_range() might make more sense here.
Thanks,
Guenter
> }
> dev_err(dev, "FIFO reset timed out\n");
>
>
On Fri, 2020-07-24 at 16:45 +0930, Joel Stanley wrote:
> From: Joachim Fenkes <[email protected]>
>
> When the SBE requests a reset via the down FIFO, that is also the
> FIFO we should go and reset ;)
Is it ?
I no longer work for IBM and dont have access to any of the
documentation here but I had vague memories that we would get a reset
request in the down fifo in order to reset the up one since we control
the up one and the host controls the down one, no ?
Cheers,
Ben.
> Fixes: 9f4a8a2d7f9d ("fsi/sbefifo: Add driver for the SBE FIFO")
> Signed-off-by: Joachim Fenkes <[email protected]>
> Signed-off-by: Joel Stanley <[email protected]>
> ---
> drivers/fsi/fsi-sbefifo.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/fsi/fsi-sbefifo.c b/drivers/fsi/fsi-sbefifo.c
> index f54df9ebc8b3..655b45c1f6ba 100644
> --- a/drivers/fsi/fsi-sbefifo.c
> +++ b/drivers/fsi/fsi-sbefifo.c
> @@ -400,7 +400,7 @@ static int sbefifo_cleanup_hw(struct sbefifo
> *sbefifo)
> /* The FIFO already contains a reset request from the SBE ? */
> if (down_status & SBEFIFO_STS_RESET_REQ) {
> dev_info(dev, "Cleanup: FIFO reset request set,
> resetting\n");
> - rc = sbefifo_regw(sbefifo, SBEFIFO_UP,
> SBEFIFO_PERFORM_RESET);
> + rc = sbefifo_regw(sbefifo, SBEFIFO_DOWN,
> SBEFIFO_PERFORM_RESET);
> if (rc) {
> sbefifo->broken = true;
> dev_err(dev, "Cleanup: Reset reg write failed,
> rc=%d\n", rc);
On Fri, 24 Jul 2020 at 14:34, Guenter Roeck <[email protected]> wrote:
>
> On 7/24/20 12:15 AM, Joel Stanley wrote:
> > From: Joachim Fenkes <[email protected]>
> >
> > On BMCs with lower timer resolution than 1ms, msleep(1) will take
> > way longer than 1ms, so looping 10k times won't wait for 10s but
> > significantly longer.
> >
> > Fix this by using jiffies like the rest of the code.
> >
> > Fixes: 9f4a8a2d7f9d ("fsi/sbefifo: Add driver for the SBE FIFO")
> > Signed-off-by: Joachim Fenkes <[email protected]>
> > Signed-off-by: Joel Stanley <[email protected]>
> > ---
> > drivers/fsi/fsi-sbefifo.c | 6 ++++--
> > 1 file changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/fsi/fsi-sbefifo.c b/drivers/fsi/fsi-sbefifo.c
> > index 655b45c1f6ba..3ad9510ad4a4 100644
> > --- a/drivers/fsi/fsi-sbefifo.c
> > +++ b/drivers/fsi/fsi-sbefifo.c
> > @@ -325,6 +325,7 @@ static int sbefifo_up_write(struct sbefifo *sbefifo, __be32 word)
> > static int sbefifo_request_reset(struct sbefifo *sbefifo)
> > {
> > struct device *dev = &sbefifo->fsi_dev->dev;
> > + unsigned long end_time;
> > u32 status, timeout;
> > int rc;
> >
> > @@ -341,7 +342,8 @@ static int sbefifo_request_reset(struct sbefifo *sbefifo)
> > }
> >
> > /* Wait for it to complete */
> > - for (timeout = 0; timeout < SBEFIFO_RESET_TIMEOUT; timeout++) {
> > + end_time = jiffies + msecs_to_jiffies(SBEFIFO_RESET_TIMEOUT);
> > + while (!time_after(jiffies, end_time)) {
> > rc = sbefifo_regr(sbefifo, SBEFIFO_UP | SBEFIFO_STS, &status);
> > if (rc) {
> > dev_err(dev, "Failed to read UP fifo status during reset"
> > @@ -355,7 +357,7 @@ static int sbefifo_request_reset(struct sbefifo *sbefifo)
> > return 0;
> > }
> >
> > - msleep(1);
> > + cond_resched();
>
> A hot loop ? Are you sure ? usleep_range() might make more sense here.
Thanks for the review.
Joachim, I can fix this up for you.
Do you have a suggestion for how long to wait? What's the best case
completion time for a sbe reset?
We also have a hot loop in sbefifo_wait. I'm open to suggestions for
how long that should delay for.
Cheers,
Joel
On Mon, 27 Jul 2020 at 05:51, Benjamin Herrenschmidt
<[email protected]> wrote:
>
> On Fri, 2020-07-24 at 16:45 +0930, Joel Stanley wrote:
> > From: Joachim Fenkes <[email protected]>
> >
> > When the SBE requests a reset via the down FIFO, that is also the
> > FIFO we should go and reset ;)
>
> Is it ?
>
> I no longer work for IBM and dont have access to any of the
> documentation here but I had vague memories that we would get a reset
> request in the down fifo in order to reset the up one since we control
> the up one and the host controls the down one, no ?
Joachim, can you clarify the situation here?
>
> > Fixes: 9f4a8a2d7f9d ("fsi/sbefifo: Add driver for the SBE FIFO")
> > Signed-off-by: Joachim Fenkes <[email protected]>
> > Signed-off-by: Joel Stanley <[email protected]>
> > ---
> > drivers/fsi/fsi-sbefifo.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/fsi/fsi-sbefifo.c b/drivers/fsi/fsi-sbefifo.c
> > index f54df9ebc8b3..655b45c1f6ba 100644
> > --- a/drivers/fsi/fsi-sbefifo.c
> > +++ b/drivers/fsi/fsi-sbefifo.c
> > @@ -400,7 +400,7 @@ static int sbefifo_cleanup_hw(struct sbefifo
> > *sbefifo)
> > /* The FIFO already contains a reset request from the SBE ? */
> > if (down_status & SBEFIFO_STS_RESET_REQ) {
> > dev_info(dev, "Cleanup: FIFO reset request set,
> > resetting\n");
> > - rc = sbefifo_regw(sbefifo, SBEFIFO_UP,
> > SBEFIFO_PERFORM_RESET);
> > + rc = sbefifo_regw(sbefifo, SBEFIFO_DOWN,
> > SBEFIFO_PERFORM_RESET);
> > if (rc) {
> > sbefifo->broken = true;
> > dev_err(dev, "Cleanup: Reset reg write failed,
> > rc=%d\n", rc);
>