diff options
author | David S. Miller <davem@davemloft.net> | 2011-11-15 12:57:00 -0800 |
---|---|---|
committer | David S. Miller <davem@davemloft.net> | 2011-11-15 12:57:00 -0800 |
commit | e88d2468718b0789b4c33da2f7e1cef2a1eee279 (patch) | |
tree | beb5a9eff292f1faa05eb9b96b7f1bd7815da131 /arch/sparc/mm/Makefile | |
parent | 1d299bc7732c34d85bd43ac1a8745f5a2fed2078 (diff) | |
download | lwn-e88d2468718b0789b4c33da2f7e1cef2a1eee279.tar.gz lwn-e88d2468718b0789b4c33da2f7e1cef2a1eee279.zip |
sparc: Stash orig_i0 into %g6 instead of %g2
As per the comments added by this commit, %g2 turns out to not be a
usable place to save away orig_i0 for syscall restart handling.
In fact all of %g2, %g3, %g4, and %g5 are assumed to be saved across
a system call by various bits of code in glibc.
%g1 can't be used because that holds the syscall number, which would
need to be saved and restored for syscall restart handling too, and
that would only compound our problems :-)
This leaves us with %g6 and %g7 which are for "system use". %g7 is
used as the "thread register" by glibc, but %g6 is used as a compiler
and assembler temporary scratch register. And in no instance is %g6
used to hold a value across a system call.
Therefore %g6 is safe for storing away orig_i0, at least for now.
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'arch/sparc/mm/Makefile')
0 files changed, 0 insertions, 0 deletions