Skip to contentSkip to conversation

patch series moving compat syscalls from fs/compat.c

DISCUSSION Stephen Rothwell 2 participants 3 messages Apr 24, 2017

lore
Share & tools
lore
#0 lore raw

Hi Al,

I couldn't easily find this patch series posted anywhere (I didn't try too hard) so I sill just comment here. I first noticed them in the vfs tree in linux-next today.

Overall, I like what they day, but when I first created kernel/compat.c (in 2002), Linus did not want the compat code sprinkled all over the other files (due to the uglifiying effect of #ifdefs in the C code, and because the compat layer was only secondary). He may have changed his mind since then, but it is worth asking. The cleanups (including more static functions) make it worth while for me, at least.

It also might have been worth asking the original (and subsequent) authors of the code for review and/or comments.

One thing I noticed from a quick look: do_handle_open() could now be static.

Signature · 2 lines
Cheers,
Stephen Rothwell
#1 lore raw
Quoted 13 lines · Stephen Rothwell

Hi Al,

I couldn't easily find this patch series posted anywhere (I didn't try too hard) so I sill just comment here. I first noticed them in the vfs tree in linux-next today.

Overall, I like what they day, but when I first created kernel/compat.c (in 2002), Linus did not want the compat code sprinkled all over the other files (due to the uglifiying effect of #ifdefs in the C code, and because the compat layer was only secondary). He may have changed his mind since then, but it is worth asking. The cleanups (including more static functions) make it worth while for me, at least.

Quoted 2 lines

It also might have been worth asking the original (and subsequent) authors of the code for review and/or comments.

FWIW, the same kind of stuff had been done before (e.g. compat
variants of readv/writev moved to fs/read_write.c, compat aio syscalls to fs/aio.c, etc.) with no objections from anybody.

The situation with kernel/compat.c is slightly different, but fs/compat.c contains very little shared infrastructure - almost all of it is straight "here's compat variant of this syscall, making use of the guts of the native one".

#2 lore raw

Hi Al,

Quoted 4 lines · On Mon, 24 Apr 2017 01:55:30 +0100 Al Viro <[email protected]> wrote:

FWIW, the same kind of stuff had been done before (e.g. compat
variants of readv/writev moved to fs/read_write.c, compat aio syscalls to fs/aio.c, etc.) with no objections from anybody.

OK. I guess I was just suprised by this series out of nowhere.

Signature · 2 lines
Cheers,
Stephen Rothwell