Re: [PATCH 3/6] linux-user: Update SO_TIMESTAMP to SO_TIMESTAMP_OLD/NEW

[Laurent Vivier]

Le 23/07/2020 à 02:19, Shu-Chun Weng a écrit :
> Both guest options map to host SO_TIMESTAMP while keeping a bit in
> fd_trans to remember if the guest expects the old or the new format.

I don't think we need to keep this information for each fd.

Once a program has used the _NEW version it will always use the _NEW
version. It's possible to mix, but I don't think we have to support
this. This adds too much complexity.

> Added a multiarch test to verify.
> 
> Signed-off-by: Shu-Chun Weng <scw@google.com>
> ---
>  linux-user/alpha/sockbits.h            |   8 +-
>  linux-user/fd-trans.h                  |  41 +++-
>  linux-user/generic/sockbits.h          |   9 +-
>  linux-user/hppa/sockbits.h             |   8 +-
>  linux-user/mips/sockbits.h             |   8 +-
>  linux-user/sparc/sockbits.h            |   8 +-
>  linux-user/strace.c                    |   7 +-(fd)
>  linux-user/syscall.c                   |  69 ++++--
>  tests/tcg/multiarch/socket_timestamp.c | 292 +++++++++++++++++++++++++
>  9 files changed, 419 insertions(+), 31 deletions(-)
>  create mode 100644 tests/tcg/multiarch/socket_timestamp.c
> 

Thanks,
Laurent