Re: [PATCH 1/2][next] scsi: qla2xxx: Replace one-element array with DECLARE_FLEX_ARRAY() helper

From: Christophe JAILLET
Date: Sat Nov 19 2022 - 05:31:15 EST


Le 19/11/2022 à 09:56, Gustavo A. R. Silva a écrit :
On Sat, Nov 19, 2022 at 09:44:02AM +0100, Christophe JAILLET wrote:
Le 19/11/2022 à 00:47, Gustavo A. R. Silva a écrit :
One-element arrays as fake flex arrays are deprecated and we are moving
towards adopting C99 flexible-array members, instead. So, replace
one-element array declaration in struct ct_sns_gpnft_rsp, which is
ultimately being used inside a union:

drivers/scsi/qla2xxx/qla_def.h:
3240 struct ct_sns_gpnft_pkt {
3241 union {
3242 struct ct_sns_req req;
3243 struct ct_sns_gpnft_rsp rsp;
3244 } p;
3245 };

Important to mention is that doing a build before/after this patch results
in no binary differences.

Hi,

even with the:

rspsz = sizeof(struct ct_sns_gpnft_rsp) +
- ((vha->hw->max_fibre_devices - 1) *
+ (vha->hw->max_fibre_devices *
sizeof(struct ct_sns_gpn_ft_data));

change ?

Yep; that change compensates for the removal of the 1 in the declaration
of entries[].

The above piece of code is a common idiom to calculate the size for an
allocation when a one-element array is involved. In the original code
(vha->hw->max_fibre_devices - 1) compensates for the _extra_ size of one
element of type struct ct_sns_gpn_ft_data in sizeof(struct ct_sns_gpnft_rsp).


Yes, I do agree, that the code is equivalent. I was surprised that a compiler was smart enough to generate the same binary code.

With gcc 11.3.0 (x86_64), I do get some differences when I do:
make drivers/scsi/qla2xxx/qla_gs.o
objdump -D drivers/scsi/qla2xxx/qla_gs.o > before.asm

patch -p1 < patch.diff

make drivers/scsi/qla2xxx/qla_gs.o
objdump -D drivers/scsi/qla2xxx/qla_gs.o > after.asm

diff -u before.asm after.asm

Mostly some slight ordering of instruction changes, but the binary are not the same.

CJ

--
Gustavo


CJ


This help us make progress towards globally enabling
-fstrict-flex-arrays=3 [1].

Link: https://github.com/KSPP/linux/issues/245
Link: https://github.com/KSPP/linux/issues/193
Link: https://gcc.gnu.org/pipermail/gcc-patches/2022-October/602902.html [1]
Signed-off-by: Gustavo A. R. Silva <gustavoars@xxxxxxxxxx>
---
drivers/scsi/qla2xxx/qla_def.h | 4 ++--
drivers/scsi/qla2xxx/qla_gs.c | 2 +-
2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h
index a26a373be9da..1eea977ef426 100644
--- a/drivers/scsi/qla2xxx/qla_def.h
+++ b/drivers/scsi/qla2xxx/qla_def.h
@@ -3151,12 +3151,12 @@ struct ct_sns_gpnft_rsp {
uint8_t vendor_unique;
};
/* Assume the largest number of targets for the union */
- struct ct_sns_gpn_ft_data {
+ DECLARE_FLEX_ARRAY(struct ct_sns_gpn_ft_data {
u8 control_byte;
u8 port_id[3];
u32 reserved;
u8 port_name[8];
- } entries[1];
+ }, entries);
};
/* CT command response */
diff --git a/drivers/scsi/qla2xxx/qla_gs.c b/drivers/scsi/qla2xxx/qla_gs.c
index 64ab070b8716..69d3bc795f90 100644
--- a/drivers/scsi/qla2xxx/qla_gs.c
+++ b/drivers/scsi/qla2xxx/qla_gs.c
@@ -4073,7 +4073,7 @@ int qla24xx_async_gpnft(scsi_qla_host_t *vha, u8 fc4_type, srb_t *sp)
sp->u.iocb_cmd.u.ctarg.req_size = GPN_FT_REQ_SIZE;
rspsz = sizeof(struct ct_sns_gpnft_rsp) +
- ((vha->hw->max_fibre_devices - 1) *
+ (vha->hw->max_fibre_devices *
sizeof(struct ct_sns_gpn_ft_data));
sp->u.iocb_cmd.u.ctarg.rsp = dma_alloc_coherent(&vha->hw->pdev->dev,