Skip to contentSkip to conversation

mmc: mmc_spi: Simplify busy loop in mmc_spi_skip()

PATCH In review Andy Shevchenko 3 participants 4 messages Jun 23, 2021 – Jul 8, 2021

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mmc: mmc_spi: Simplify busy loop in mmc_spi_skip()

Infinite loops are hard to read and understand because of hidden main loop condition. Simplify such one in mmc_spi_skip().

Using schedule() to schedule (and be friendly to others) is discouraged and cond_resched() should be used instead. Hence, replace schedule() with cond_resched() at the same time.

---
 drivers/mmc/host/mmc_spi.c | 15 ++++-----------
 1 file changed, 4 insertions(+), 11 deletions(-)
drivers/mmc/host/mmc_spi.c +4 -11
index 65c65bb5737f..a1bcde3395a6 100644
@@ -180,7 +180,7 @@ static int mmc_spi_skip(struct mmc_spi_host *host, unsigned long timeout,
 	u8 *cp = host->data->status;
 	unsigned long start = jiffies;
 
-	while (1) {
+	do {
 		int		status;
 		unsigned	i;
 
@@ -193,16 +193,9 @@ static int mmc_spi_skip(struct mmc_spi_host *host, unsigned long timeout,
 				return cp[i];
 		}
 
-		if (time_is_before_jiffies(start + timeout))
-			break;
-
-		/* If we need long timeouts, we may release the CPU.
-		 * We use jiffies here because we want to have a relation
-		 * between elapsed time and the blocking of the scheduler.
-		 */
-		if (time_is_before_jiffies(start + 1))
-			schedule();
-	}
+		/* If we need long timeouts, we may release the CPU */
+		cond_resched();
+	} while (time_is_after_jiffies(start + timeout));
 	return -ETIMEDOUT;
 }
 
Signature · 1 line
2.30.2
#1 lore raw
Quoted 34 lines · Andy Shevchenko

Infinite loops are hard to read and understand because of hidden main loop condition. Simplify such one in mmc_spi_skip().

Using schedule() to schedule (and be friendly to others) is discouraged and cond_resched() should be used instead. Hence, replace schedule() with cond_resched() at the same time.

---
 drivers/mmc/host/mmc_spi.c | 15 ++++-----------
 1 file changed, 4 insertions(+), 11 deletions(-)
drivers/mmc/host/mmc_spi.c +1 -4
index 65c65bb5737f..a1bcde3395a6 100644
@@ -180,7 +180,7 @@ static int mmc_spi_skip(struct mmc_spi_host *host, unsigned long timeout,
        u8 *cp = host->data->status;
        unsigned long start = jiffies;
 
-       while (1) {
+       do {
                int             status;
                unsigned        i;
 
@@ -193,16 +193,9 @@ static int mmc_spi_skip(struct mmc_spi_host *host, unsigned long timeout,
                                return cp[i];
                }
 
-               if (time_is_before_jiffies(start + timeout))
-                       break;
-

drivers/mmc/host/mmc_spi.c static int mmc_spi_skip(struct mmc_spi_host *host, unsigned long timeout, · line 198

Quoted 10 lines
drivers/mmc/host/mmc_spi.c +3 -7
-               /* If we need long timeouts, we may release the CPU.
-                * We use jiffies here because we want to have a relation
-                * between elapsed time and the blocking of the scheduler.
-                */
-               if (time_is_before_jiffies(start + 1))
-                       schedule();
-       }
+               /* If we need long timeouts, we may release the CPU */
+               cond_resched();
+       } while (time_is_after_jiffies(start + timeout));

This certainly is an improvement.

Although, what do you think of moving to readx_poll_timeout(), that should allow even a better cleanup, don't you think?

drivers/mmc/host/mmc_spi.c static int mmc_spi_skip(struct mmc_spi_host *host, unsigned long timeout, · line 201

Quoted 3 lines
drivers/mmc/host/mmc_spi.c +0 -0
        return -ETIMEDOUT;
 }
 

Kind regards
Uffe

#2 lore raw
Quoted 2 lines · Ulf Hansson

On Wed, 23 Jun 2021 at 12:17, Andy Shevchenko
<[email protected]> wrote:

...

Quoted 4 lines

This certainly is an improvement.

Although, what do you think of moving to readx_poll_timeout(), that should allow even a better cleanup, don't you think?

I believe you meant rather read_poll_timeout(). Either way I don't see the benefit of using that macro when you have to customize its body a lot. Besides that the macro doesn't use cond_sched() or even schedule() and I'm not sure it will be an equivalent change.

That said, I prefer going this patch as is for the time being. We may adjust it later on.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#3 lore raw
Quoted 12 lines · Andy Shevchenko
Quoted 2 lines · Ulf Hansson

On Wed, 23 Jun 2021 at 12:17, Andy Shevchenko
<[email protected]> wrote:

...

Quoted 4 lines

This certainly is an improvement.

Although, what do you think of moving to readx_poll_timeout(), that should allow even a better cleanup, don't you think?

Quoted 7 lines

I believe you meant rather read_poll_timeout(). Either way I don't see the benefit of using that macro when you have to customize its body a lot. Besides that the macro doesn't use cond_sched() or even schedule() and I'm not sure it will be an equivalent change.

That said, I prefer going this patch as is for the time being. We may adjust it later on.

Okay, no strong opinion from my side. Queued for v5.15 on my devel branch, thanks!

Kind regards
Uffe