[patch] ARM ld.so broken

[Paul Brook]

The patch below fixes a bug in the new ARM _ld_linux_resolve implementation.
I'm don't know if/how the current implementation was tested, but it's 
completely broken.

This patch makes the prologue and epilogue agree on how big the stack frame 
is, and also makes sure EABI doubleword stack alignment is preserved.

Tested on arm-linux-gnueabi.

Paul

--- ldso/ldso/arm/resolve.S	(revision 154935)
+++ ldso/ldso/arm/resolve.S	(local)
@@ -108,8 +108,10 @@ _dl_linux_resolve:
          @ function must branch to the real function, and that expects
          @ r0-r3 and lr to be as they were before the whole PLT stuff -
          @ ip can be trashed.
+	 @ This routine is called after pushing lr, so we must push an odd
+	 @ number of words to keep the stack correctly aligned.
 
-         stmdb sp!, {r0, r1, r2, r3, sl, fp}
+         stmdb sp!, {r0, r1, r2, r3, r4}
          ldr r0, [lr, #-4]       @ r0 :        = [lr-4] (GOT_TABLE[1])
          sub r1, lr, ip          @ r1 :        = (lr-ip) (a multple of 4)
          mvn r1, r1, ASR #2      @ r1 :        = ~((lr-ip)>>2), since -x = 
(1+~x)
@@ -119,7 +121,7 @@ _dl_linux_resolve:
 	bl _dl_linux_resolver
 
 	mov ip, r0
-        ldmia sp!, {r0-r3, lr}
+        ldmia sp!, {r0, r1, r2, r3, r4, lr}
 
 #if defined(__USE_BX__)
 	bx ip
@@ -141,7 +143,9 @@ _dl_linux_resolve:
        @ function must branch to the real function, and that expects
        @ r0-r3 and lr to be as they were before the whole PLT stuff -
        @ ip can be trashed.
-       push    {r0-r3}
+       @ This routine is called after pushing lr, so we must push an odd
+       @ number of words to keep the stack correctly aligned.
+       push    {r0-r4}
        mov     r1, lr          @ &GOT_TABLE[2]
        sub     r0, r1, #4
        mov     r2, ip          @ &GOT[n]
@@ -154,9 +158,9 @@ _dl_linux_resolve:
        @ r0 contains the branch address, the return address is above
        @ the saved r0..r3
        mov     ip, r0
-       ldr     r1, [sp, #16]
+       ldr     r1, [sp, #20]
        mov     lr, r1
-       pop     {r0-r3}
+       pop     {r0-r4}
        add     sp, #4
        bx      ip