[PATCH] RDMA/irdma: fix irdma_alloc_ucontext_resp padding

[PATCH] RDMA/irdma: fix irdma_alloc_ucontext_resp padding

[Arnd Bergmann]

From: Arnd Bergmann <arnd@arndb.de>

A recent modified struct irdma_alloc_ucontext_resp by adding a member
with implicit padding in front of it, changing the ABI in an
incompatibible way on all architectures other than m68k, as
reported by scripts/check-uapi.sh:

==== ABI differences detected in include/rdma/irdma-abi.h from 1dd7bde2e91c -> HEAD ====
    [C] 'struct irdma_alloc_ucontext_resp' changed:
      type size changed from 704 to 640 (in bits)
      1 data member deletion:
        '__u8 rsvd3[2]', at offset 640 (in bits) at irdma-abi.h:61:1
      1 data member insertion:
        '__u8 revd3[2]', at offset 592 (in bits) at irdma-abi.h:60:1

Change the ABI back to the previous version, by moving the new
max_hw_srq_quanta member into a naturally aligned location.

Fixes: 563e1feb5f6e ("RDMA/irdma: Add SRQ support")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 include/uapi/rdma/irdma-abi.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/uapi/rdma/irdma-abi.h b/include/uapi/rdma/irdma-abi.h
index f7788d33376b..36f20802bcc8 100644
--- a/include/uapi/rdma/irdma-abi.h
+++ b/include/uapi/rdma/irdma-abi.h
@@ -57,8 +57,8 @@ struct irdma_alloc_ucontext_resp {
 	__u8 rsvd2;
 	__aligned_u64 comp_mask;
 	__u16 min_hw_wq_size;
+	__u8 revd3[2];
 	__u32 max_hw_srq_quanta;
-	__u8 rsvd3[2];
 };
 
 struct irdma_alloc_pd_resp {
-- 
2.39.5

Re: [PATCH] RDMA/irdma: fix irdma_alloc_ucontext_resp padding

[Geert Uytterhoeven]

Hi Arnd,

On Mon, 8 Dec 2025 at 14:39, Arnd Bergmann <arnd@kernel.org> wrote:
> From: Arnd Bergmann <arnd@arndb.de>
>
> A recent modified struct irdma_alloc_ucontext_resp by adding a member
> with implicit padding in front of it, changing the ABI in an
> incompatibible way on all architectures other than m68k, as
> reported by scripts/check-uapi.sh:
>
> ==== ABI differences detected in include/rdma/irdma-abi.h from 1dd7bde2e91c -> HEAD ====
>     [C] 'struct irdma_alloc_ucontext_resp' changed:
>       type size changed from 704 to 640 (in bits)
>       1 data member deletion:
>         '__u8 rsvd3[2]', at offset 640 (in bits) at irdma-abi.h:61:1
>       1 data member insertion:
>         '__u8 revd3[2]', at offset 592 (in bits) at irdma-abi.h:60:1
>
> Change the ABI back to the previous version, by moving the new
> max_hw_srq_quanta member into a naturally aligned location.
>
> Fixes: 563e1feb5f6e ("RDMA/irdma: Add SRQ support")
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Thanks for the discussion in Tokyo!

Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH] RDMA/irdma: fix irdma_alloc_ucontext_resp padding

[Jacob Moroni]

Hi,

This doesn't change the offset of max_hw_srq_quanta on my system, but I tested
with a verbs provider built with the previous and new proposed change
just in case,
and both worked.

Out of curiosity, what is the policy for increasing the size of these
uverbs driver
response structures? I think the response gets silently truncated if the user
provides a smaller buffer, so this shouldn't have broken any user applications
using the old ABI IIUC.

Thanks!

Tested-by: Jacob Moroni <jmoroni@google.com>

On Mon, Dec 15, 2025 at 1:43 AM Geert Uytterhoeven <geert@linux-m68k.org> wrote:
>
> Hi Arnd,
>
> On Mon, 8 Dec 2025 at 14:39, Arnd Bergmann <arnd@kernel.org> wrote:
> > From: Arnd Bergmann <arnd@arndb.de>
> >
> > A recent modified struct irdma_alloc_ucontext_resp by adding a member
> > with implicit padding in front of it, changing the ABI in an
> > incompatibible way on all architectures other than m68k, as
> > reported by scripts/check-uapi.sh:
> >
> > ==== ABI differences detected in include/rdma/irdma-abi.h from 1dd7bde2e91c -> HEAD ====
> >     [C] 'struct irdma_alloc_ucontext_resp' changed:
> >       type size changed from 704 to 640 (in bits)
> >       1 data member deletion:
> >         '__u8 rsvd3[2]', at offset 640 (in bits) at irdma-abi.h:61:1
> >       1 data member insertion:
> >         '__u8 revd3[2]', at offset 592 (in bits) at irdma-abi.h:60:1
> >
> > Change the ABI back to the previous version, by moving the new
> > max_hw_srq_quanta member into a naturally aligned location.
> >
> > Fixes: 563e1feb5f6e ("RDMA/irdma: Add SRQ support")
> > Signed-off-by: Arnd Bergmann <arnd@arndb.de>
>
> Thanks for the discussion in Tokyo!
>
> Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
>
> Gr{oetje,eeting}s,
>
>                         Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
>
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds
>

Re: [PATCH] RDMA/irdma: fix irdma_alloc_ucontext_resp padding

[Geert Uytterhoeven]

Hi Jacob,

On Tue, 16 Dec 2025 at 15:17, Jacob Moroni <jmoroni@google.com> wrote:
> This doesn't change the offset of max_hw_srq_quanta on my system, but I tested
> with a verbs provider built with the previous and new proposed change
> just in case,
> and both worked.

It indeed doesn't change the offset of that field (except on m68k),
but it should restore the size of the structure to what it was before.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH] RDMA/irdma: fix irdma_alloc_ucontext_resp padding

[Jason Gunthorpe]

On Mon, Dec 08, 2025 at 02:38:44PM +0100, Arnd Bergmann wrote:
> From: Arnd Bergmann <arnd@arndb.de>
> 
> A recent modified struct irdma_alloc_ucontext_resp by adding a member
> with implicit padding in front of it, changing the ABI in an
> incompatibible way on all architectures other than m68k, as
> reported by scripts/check-uapi.sh:

The growth of the structure should be tolerated, that is how the ABI
is expected to work, though this is wrong and should be fixed while we
still can.

Applied to for-rc

Thanks,
Jason