[PATCH 1/3] scsi: hisi_sas: Use devm_bitmap_zalloc() when applicable

[PATCH 1/3] scsi: hisi_sas: Use devm_bitmap_zalloc() when applicable

[Christophe JAILLET]

'hisi_hba->slot_index_tags' is a bitmap. So use 'devm_bitmap_zalloc()' to
simplify code, improve the semantic and avoid some open-coded arithmetic
in allocator arguments.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
The use of 's' is questionable here. I've left it because it looks more
consistent this way with the surrounding code.

Can it be an issue to have the length of the allocated bitmap not being
a multiple of sizeof(long)?
I guess that there is some kind of 'rounding' done by the memory allocator
to keep some alignment, so I think that the previous code is safe (but not
logical).
If this is not the case, there is a potential out of bound bug related to
the bitmap API that expect to access only longs (which is not necessarily
the case here).
---
 drivers/scsi/hisi_sas/hisi_sas_main.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/scsi/hisi_sas/hisi_sas_main.c b/drivers/scsi/hisi_sas/hisi_sas_main.c
index f206c433de32..6ecb42d5ce81 100644
--- a/drivers/scsi/hisi_sas/hisi_sas_main.c
+++ b/drivers/scsi/hisi_sas/hisi_sas_main.c
@@ -2516,9 +2516,8 @@ int hisi_sas_alloc(struct hisi_hba *hisi_hba)
 	if (!hisi_hba->breakpoint)
 		goto err_out;
 
-	hisi_hba->slot_index_count = max_command_entries;
-	s = hisi_hba->slot_index_count / BITS_PER_BYTE;
-	hisi_hba->slot_index_tags = devm_kzalloc(dev, s, GFP_KERNEL);
+	s = hisi_hba->slot_index_count = max_command_entries;
+	hisi_hba->slot_index_tags = devm_bitmap_zalloc(dev, s, GFP_KERNEL);
 	if (!hisi_hba->slot_index_tags)
 		goto err_out;
 
-- 
2.30.2

[PATCH 2/3] scsi: hisi_sas: Remove some useless code in 'hisi_sas_alloc()'

[Christophe JAILLET]

The 'hisi_hba->slot_index_tags' bitmap is allocated with 'bitmap_zalloc()'
so it is already cleared.

There is no need to clear it another time, one bit at a time.

So, remove the corresponding useless code.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
 drivers/scsi/hisi_sas/hisi_sas_main.c | 9 ---------
 1 file changed, 9 deletions(-)

diff --git a/drivers/scsi/hisi_sas/hisi_sas_main.c b/drivers/scsi/hisi_sas/hisi_sas_main.c
index 6ecb42d5ce81..d4f5d093bde4 100644
--- a/drivers/scsi/hisi_sas/hisi_sas_main.c
+++ b/drivers/scsi/hisi_sas/hisi_sas_main.c
@@ -206,14 +206,6 @@ static int hisi_sas_slot_index_alloc(struct hisi_hba *hisi_hba,
 	return index;
 }
 
-static void hisi_sas_slot_index_init(struct hisi_hba *hisi_hba)
-{
-	int i;
-
-	for (i = 0; i < hisi_hba->slot_index_count; ++i)
-		hisi_sas_slot_index_clear(hisi_hba, i);
-}
-
 void hisi_sas_slot_task_free(struct hisi_hba *hisi_hba, struct sas_task *task,
 			     struct hisi_sas_slot *slot)
 {
@@ -2535,7 +2527,6 @@ int hisi_sas_alloc(struct hisi_hba *hisi_hba)
 	if (!hisi_hba->sata_breakpoint)
 		goto err_out;
 
-	hisi_sas_slot_index_init(hisi_hba);
 	hisi_hba->last_slot_index = HISI_SAS_UNRESERVED_IPTT;
 
 	hisi_hba->wq = create_singlethread_workqueue(dev_name(dev));
-- 
2.30.2

Re: [PATCH 2/3] scsi: hisi_sas: Remove some useless code in 'hisi_sas_alloc()'

[John Garry]

On 26/11/2021 21:18, Christophe JAILLET wrote:
> The 'hisi_hba->slot_index_tags' bitmap is allocated with 'bitmap_zalloc()'
> so it is already cleared.
> 
> There is no need to clear it another time, one bit at a time.
> 
> So, remove the corresponding useless code.
> 
> Signed-off-by: Christophe JAILLET<christophe.jaillet@wanadoo.fr>

Acked-by: John Garry <john.garry@huawei.com>

[PATCH 3/3] scsi: hisi_sas: Use non-atomic bitmap functions when possible

[Christophe JAILLET]

All uses of the 'hisi_hba->slot_index_tags' bitmap are protected with the
'hisi_hba->lock' spinlock.

So prefer the non-atomic '__[set|clear]_bit()' functions to save a few
cycles.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
 drivers/scsi/hisi_sas/hisi_sas_main.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/scsi/hisi_sas/hisi_sas_main.c b/drivers/scsi/hisi_sas/hisi_sas_main.c
index d4f5d093bde4..889c36fa9309 100644
--- a/drivers/scsi/hisi_sas/hisi_sas_main.c
+++ b/drivers/scsi/hisi_sas/hisi_sas_main.c
@@ -158,7 +158,7 @@ static void hisi_sas_slot_index_clear(struct hisi_hba *hisi_hba, int slot_idx)
 {
 	void *bitmap = hisi_hba->slot_index_tags;
 
-	clear_bit(slot_idx, bitmap);
+	__clear_bit(slot_idx, bitmap);
 }
 
 static void hisi_sas_slot_index_free(struct hisi_hba *hisi_hba, int slot_idx)
@@ -175,7 +175,7 @@ static void hisi_sas_slot_index_set(struct hisi_hba *hisi_hba, int slot_idx)
 {
 	void *bitmap = hisi_hba->slot_index_tags;
 
-	set_bit(slot_idx, bitmap);
+	__set_bit(slot_idx, bitmap);
 }
 
 static int hisi_sas_slot_index_alloc(struct hisi_hba *hisi_hba,
-- 
2.30.2

Re: [PATCH 3/3] scsi: hisi_sas: Use non-atomic bitmap functions when possible

[John Garry]

On 26/11/2021 21:18, Christophe JAILLET wrote:
> All uses of the 'hisi_hba->slot_index_tags' bitmap are protected with the
> 'hisi_hba->lock' spinlock.
> 
> So prefer the non-atomic '__[set|clear]_bit()' functions to save a few
> cycles.
> 
> Signed-off-by: Christophe JAILLET<christophe.jaillet@wanadoo.fr>

Acked-by: John Garry <john.garry@huawei.com>

Re: [PATCH 1/3] scsi: hisi_sas: Use devm_bitmap_zalloc() when applicable

[John Garry]

On 26/11/2021 21:15, Christophe JAILLET wrote:
> 'hisi_hba->slot_index_tags' is a bitmap. So use 'devm_bitmap_zalloc()' to
> simplify code, improve the semantic and avoid some open-coded arithmetic
> in allocator arguments.
> 
> Signed-off-by: Christophe JAILLET<christophe.jaillet@wanadoo.fr>

Acked-by: John Garry <john.garry@huawei.com>

> ---
> The use of 's' is questionable here. I've left it because it looks more
> consistent this way with the surrounding code.
> 
> Can it be an issue to have the length of the allocated bitmap not being
> a multiple of sizeof(long)?

The driver does not rely on that (allocated bitmap being a multiple of 
sizeof(long)), but the size is 4096 bits, which would be a multiple of 
sizeof(long)

Thanks,
John

> I guess that there is some kind of 'rounding' done by the memory allocator
> to keep some alignment, so I think that the previous code is safe (but not
> logical).
> If this is not the case, there is a potential out of bound bug related to
> the bitmap API that expect to access only longs (which is not necessarily
> the case here).

Re: [PATCH 1/3] scsi: hisi_sas: Use devm_bitmap_zalloc() when applicable

[Martin K. Petersen]

Christophe,

> 'hisi_hba->slot_index_tags' is a bitmap. So use 'devm_bitmap_zalloc()' to
> simplify code, improve the semantic and avoid some open-coded arithmetic
> in allocator arguments.

Applied to 5.17/scsi-staging, thanks!

-- 
Martin K. Petersen	Oracle Linux Engineering

Re: [PATCH 1/3] scsi: hisi_sas: Use devm_bitmap_zalloc() when applicable

[Martin K. Petersen]

On Fri, 26 Nov 2021 22:15:21 +0100, Christophe JAILLET wrote:

> 'hisi_hba->slot_index_tags' is a bitmap. So use 'devm_bitmap_zalloc()' to
> simplify code, improve the semantic and avoid some open-coded arithmetic
> in allocator arguments.
> 
> 

Applied to 5.17/scsi-queue, thanks!

[1/3] scsi: hisi_sas: Use devm_bitmap_zalloc() when applicable
      https://git.kernel.org/mkp/scsi/c/54585ec62fbd
[2/3] scsi: hisi_sas: Remove some useless code in 'hisi_sas_alloc()'
      https://git.kernel.org/mkp/scsi/c/d43efddf6271
[3/3] scsi: hisi_sas: Use non-atomic bitmap functions when possible
      https://git.kernel.org/mkp/scsi/c/4d6942e2666e

-- 
Martin K. Petersen	Oracle Linux Engineering