Re: [PATCH 5/6] FDPIC: Add coredump capability for the ELF-FDPIC binfmt [try #3]

[David Howells <dhowells@redhat.com>]

Andrew Morton <akpm@osdl.org> wrote:

> llseek takes a loff_t and file->f_pos is loff_t.  I guess it's a bit moot
> on such a CPU.  Was it deliberate?

It compiles with no error and no warning, so I haven't noticed.  This is as
binfmt_elf.c is, I believe, so that is probably wrong too.

> (how come the kernel doesn't have a SEEK_SET #define?)

I don't know.  It probably should.

> Three callsites - seems too large to inline.

Again taken from binfmt_elf.c, although I added the debugging stuff.  It
shouldn't matter as the compiler will make its own decision (or does "inline"
get #defined to always-inline nowadays?).

> Which seems reasonable to me.  I'll steal it from them.

Okay.

> Embedding returns and gotos in macros is evil.  For new code it's worth
> doing it vaguely tastefully.

Again, stolen verbatim from binfmt_elf.c.  I'd prefer to keep it comparable by
the blink-comparator method if possible.

> Does this need locking?

It shouldn't do; we own our own vma chain, and because we're part of exec, we
have a fresh mm_struct to play with.  The VMAs themselves aren't allowed to
change, not even on NOMMU.

David