block/nvme: Remove QEMU_PACKED from naturally aligned NVMeRegs struct
PATCH 1/1 block/nvme: Remove QEMU_PACKED from naturally aligned NVMeRegs struct
The QEMU_PACKED is causing a compiler warning/error with GCC 9:
CC block/nvme.oblock/nvme.c: In function ‘nvme_create_queue_pair’:block/nvme.c:209:22: error: taking address of packed member of
‘struct <anonymous>’ may result in an unaligned pointer value [-Werror=address-of-packed-member]
209 | q->sq.doorbell = &s->regs->doorbells[idx * 2 * s->doorbell_scale];
All members of the struct are naturally aligned, so there should not be the need for QEMU_PACKED here, and the following QEMU_BUILD_BUG_ON also ensures that there is no padding. Thus simply remove the QEMU_PACKED here.
--- block/nvme.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
block/nvme.c +1 -1
@@ -82,7 +82,7 @@ typedef volatile struct {uint8_t reserved1[0xec0 ];uint8_t cmd_set_specfic[0x100 ];uint32_t doorbells[]; -}QEMU_PACKED NVMeRegs; +} NVMeRegs;QEMU_BUILD_BUG_ON (offsetof(NVMeRegs , doorbells) !=0x1000 );
Signature · 1 line
1.8.3.1
Quoted 27 lines · Thomas Huth
The
QEMU_PACKEDis causing a compiler warning/error with GCC 9:CC
block/nvme.oblock/nvme.c: In function ‘nvme_create_queue_pair’:block/nvme.c:209:22: error: taking address of packed member of
‘struct <anonymous>’ may result in an unaligned pointer value [-Werror=address-of-packed-member]
209 |q->sq.doorbell= &s->regs->doorbells[idx * 2 *s->doorbell_scale];All members of the struct are naturally aligned, so there should not be the need for
QEMU_PACKEDhere, and the followingQEMU_BUILD_BUG_ONalso ensures that there is no padding. Thus simply remove theQEMU_PACKEDhere.Reported-by Satheesh Rajendran <[email protected]>Signed-off-by Thomas Huth <[email protected]>--- block/nvme.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)block/nvme.c +0
-0
block/nvme.c typedef volatile struct { · line 88
Quoted 8 lines
block/nvme.c +1
-1@@ -82,7 +82,7 @@ typedef volatile struct {uint8_t reserved1[0xec0 ];uint8_t cmd_set_specfic[0x100 ];uint32_t doorbells[]; -}QEMU_PACKED NVMeRegs; +} NVMeRegs;QEMU_BUILD_BUG_ON (offsetof(NVMeRegs , doorbells) !=0x1000 );
thanks
-- PMM
Kevin Wolf
AppliedAm 25.02.2019 um 12:59 hat Thomas Huth geschrieben:
Quoted 9 lines
The
QEMU_PACKEDis causing a compiler warning/error with GCC 9:CC
block/nvme.oblock/nvme.c: In function ‘nvme_create_queue_pair’:block/nvme.c:209:22: error: taking address of packed member of
‘struct <anonymous>’ may result in an unaligned pointer value [-Werror=address-of-packed-member]
209 |q->sq.doorbell= &s->regs->doorbells[idx * 2 *s->doorbell_scale];
Quoted 8 lines
All members of the struct are naturally aligned, so there should not be the need for
QEMU_PACKEDhere, and the followingQEMU_BUILD_BUG_ONalso ensures that there is no padding. Thus simply remove theQEMU_PACKEDhere.Reported-by Satheesh Rajendran <[email protected]>Signed-off-by Thomas Huth <[email protected]>
Thanks, applied to the block branch.
Though I'm not sure why a compiler should warn if packed and non-packed result in the exact same layout anyway...
Kevin
Quoted 2 lines · Kevin Wolf
Though I'm not sure why a compiler should warn if packed and non-packed result in the exact same layout anyway...
Packed implies "and any time you see a pointer to this struct it might not be aligned". Non-packed implies "you can assume that the base of the struct is aligned".
thanks
-- PMM
Quoted 7 lines · Peter Maydell
Quoted 2 lines · Kevin Wolf
Though I'm not sure why a compiler should warn if packed and non-packed result in the exact same layout anyway...
Packed implies "and any time you see a pointer to this struct it might not be aligned". Non-packed implies "you can assume that the base of the struct is aligned".
Right. FWIW, consider this small program:
#include <stdio.h>
#include <stdint.h>
typedef struct {
uint64_t a;
uint64_t b;
uint64_t c;
uint64_t d;
} __attribute__((packed)) s1_t;struct {
char x;
s1_t s1;
} s2;int main()
{
printf("address of s2.s1.d = %p\n", &s2.s1.d);
return 0;
}Though all members of s1_t look like they are naturally aligned, I get an odd pointer for one of its members in this case.
So passing along that pointer to other functions will certainly cause crashes on architectures like Sparc, thus the compiler warning is indeed justified here.
Thomas