2023-04-21 12:56:24

by Yahu Gao

Subject: [PATCH] drivers/ata: Improve performance and readability

From: Yahu Gao <[email protected]>

Replace conditions of avoid issuing [P]IDENTIFY to PMP.

Reviewed-by: Jiwei Sun <[email protected]>
Signed-off-by: Yahu Gao <[email protected]>

diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c
index 14c17c3bda4e..53f65d751189 100644
--- a/drivers/ata/libata-core.c
+++ b/drivers/ata/libata-core.c
@@ -3802,11 +3802,7 @@ int ata_dev_revalidate(struct ata_device *dev, unsigned int new_class,
return -ENODEV;

/* fail early if !ATA && !ATAPI to avoid issuing [P]IDENTIFY to PMP */
- if (ata_class_enabled(new_class) &&
- new_class != ATA_DEV_ATA &&
- new_class != ATA_DEV_ATAPI &&
- new_class != ATA_DEV_ZAC &&
- new_class != ATA_DEV_SEMB) {
+ if (new_class == ATA_DEV_PMP) {
ata_dev_info(dev, "class mismatch %u != %u\n",
dev->class, new_class);
rc = -ENODEV;
--
2.39.1

2023-04-21 22:14:47

by Damien Le Moal

Subject: Re: [PATCH] drivers/ata: Improve performance and readability

On 4/21/23 21:54, Yahu Gao wrote:
> From: Yahu Gao <[email protected]>
>

The patch title should start with:

ata: libata-core: ...

And let's be honest here: this is not the hot path and you are optimizing a
single "if" condition for a series of commands that will take 10s of ms to
execute. So this is *not* about performance. So update the patch title please.

> Replace conditions of avoid issuing [P]IDENTIFY to PMP.
>
> Reviewed-by: Jiwei Sun <[email protected]>
> Signed-off-by: Yahu Gao <[email protected]>
>
> diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c
> index 14c17c3bda4e..53f65d751189 100644
> --- a/drivers/ata/libata-core.c
> +++ b/drivers/ata/libata-core.c
> @@ -3802,11 +3802,7 @@ int ata_dev_revalidate(struct ata_device *dev, unsigned int new_class,
> return -ENODEV;
>
> /* fail early if !ATA && !ATAPI to avoid issuing [P]IDENTIFY to PMP */
> - if (ata_class_enabled(new_class) &&
> - new_class != ATA_DEV_ATA &&
> - new_class != ATA_DEV_ATAPI &&
> - new_class != ATA_DEV_ZAC &&
> - new_class != ATA_DEV_SEMB) {
> + if (new_class == ATA_DEV_PMP) {

Using a switch-case here may actually make things even more readable.

> ata_dev_info(dev, "class mismatch %u != %u\n",
> dev->class, new_class);
> rc = -ENODEV;