>>>>> On Tue, 19 Oct 2004 13:28:34 -0600, Erik Andersen <andersen@codepoet.org> said:
>> BTW, top of uClibc TODO list is "Fix syscall() on mips". What is a
>> problem?
andersen> It appears that uClibc's syscall() for mips works ok for
andersen> syscalls with a few arguments. But as I recall, it does not
andersen> work properly with _syscall5, _syscall6, _syscall7, etc.
andersen> Perhaps there is some mistake in its assumptions about the
andersen> mips/linux ABI regarding which syscall arguments are passed
andersen> via register vs which syscall arguments are passed on the
andersen> stack.
Hmm... I found a old fix in uClibc ML archive.
http://www.uclibc.org/lists/uclibc/2002-September/004459.html
But it seems somewhat broken. How about this fix instead? I tested
mmap with syscall() in mips. mips64 is not tested.
move a0, a1 /* Move the next three args up a register. */
move a1, a2
move a2, a3
- /* Load the remaining possible args (up to 11) from the stack. */
+ /* Load the remaining possible args (up to 7) from the stack. */
#ifdef __mips64
- ld t0,4*8(sp)
- ld t1,5*8(sp)
- ld t2,6*8(sp)
- ld t3,7*8(sp)
- ld t4,8*8(sp)
- ld t5,9*8(sp)
- ld t6,10*8(sp)
+ move a3, a4
+ move a4, a5
+ move a5, a6
+ move a6, a7
+ daddiu sp,sp,-16
+ sd v0,0*8(sp)
+ ld v0,0*8(sp) /* for system call restarts */
#else
- lw t0,4*4(sp)
- lw t1,5*4(sp)
- lw t2,6*4(sp)
- lw t3,7*4(sp)
- lw t4,8*4(sp)
- lw t5,9*4(sp)
- lw t6,10*4(sp)
+ lw a3,4*4(sp)
+ lw t0,5*4(sp)
+ lw t1,6*4(sp)
+ lw t2,7*4(sp)
+ addiu sp,sp,-32
+ sw t0,4*4(sp)
+ sw t1,5*4(sp)
+ sw t2,6*4(sp)
+ sw v0,7*4(sp)
+ lw v0,7*4(sp) /* for system call restarts */
#endif
syscall /* Do the system call. */
+#ifdef __mips64
+ daddiu sp,sp,16
+#else
+ addiu sp,sp,32
+#endif
j ra /* Return to caller. */
+.end syscall
+.size syscall,.-syscall