[PATCH, COMMITTED] BZ #15128 [AArch64] Save and restore q0-q7 on entry to dynamic linker.

[Marcus Shawcroft <marcus.shawcroft@linaro.org>]

[-- Attachment #1: Type: text/plain, Size: 229 bytes --]

This patch ensures that all argument passing registers are saved and
restored on entry to the dynamic linker and resolves BZ #15128.

Regression tested.  Committed to trunk.  Back port to 2.18 branch will follow.

Cheers
/Marcus

[-- Attachment #2: glibc-15128.txt --]
[-- Type: text/plain, Size: 4450 bytes --]

commit 18b991db8133f0cce3b61805e80dbb1e6f5e61e8
Author: Marcus Shawcroft <marcus.shawcroft@linaro.org>
Date:   Wed Dec 18 10:00:07 2013 +0000

    [AArch64] Save and restore q0-q7 on entry to dynamic linker.
    
    [BZ #15128] Ensure all argument passing registers are saved and
    restored on entry to dynamic linker.

diff --git a/NEWS b/NEWS
index 7886834..874ab33 100644
--- a/NEWS
+++ b/NEWS
@@ -12,17 +12,17 @@ Version 2.19
   156, 387, 431, 832, 926, 2801, 4772, 6786, 6787, 6807, 6810, 7003, 9954,
   10253, 10278, 11087, 11157, 11214, 12100, 12486, 13028, 13982, 13985,
   14029, 14032, 14120, 14143, 14155, 14547, 14699, 14752, 14876, 14910,
-  15004, 15048, 15089, 15218, 15268, 15277, 15308, 15362, 15374, 15400,
-  15425, 15427, 15483, 15522, 15531, 15532, 15593, 15601, 15608, 15609,
-  15610, 15632, 15640, 15670, 15672, 15680, 15681, 15723, 15734, 15735,
-  15736, 15748, 15749, 15754, 15760, 15763, 15764, 15797, 15799, 15825,
-  15843, 15844, 15847, 15849, 15855, 15856, 15857, 15859, 15867, 15886,
-  15887, 15890, 15892, 15893, 15895, 15897, 15901, 15905, 15909, 15915,
-  15917, 15919, 15921, 15923, 15939, 15941, 15948, 15963, 15966, 15985,
-  15988, 15997, 16032, 16034, 16036, 16037, 16038, 16041, 16055, 16071,
-  16072, 16074, 16077, 16078, 16103, 16112, 16143, 16144, 16146, 16150,
-  16151, 16153, 16167, 16172, 16195, 16214, 16245, 16271, 16274, 16283,
-  16289, 16314, 16316, 16330.
+  15004, 15048, 15089, 15128, 15218, 15268, 15277, 15308, 15362, 15374,
+  15400, 15425, 15427, 15483, 15522, 15531, 15532, 15593, 15601, 15608,
+  15609, 15610, 15632, 15640, 15670, 15672, 15680, 15681, 15723, 15734,
+  15735, 15736, 15748, 15749, 15754, 15760, 15763, 15764, 15797, 15799,
+  15825, 15843, 15844, 15847, 15849, 15855, 15856, 15857, 15859, 15867,
+  15886, 15887, 15890, 15892, 15893, 15895, 15897, 15901, 15905, 15909,
+  15915, 15917, 15919, 15921, 15923, 15939, 15941, 15948, 15963, 15966,
+  15985, 15988, 15997, 16032, 16034, 16036, 16037, 16038, 16041, 16055,
+  16071, 16072, 16074, 16077, 16078, 16103, 16112, 16143, 16144, 16146,
+  16150, 16151, 16153, 16167, 16172, 16195, 16214, 16245, 16271, 16274,
+  16283, 16289, 16314, 16316, 16330.
 
 * The public headers no longer use __unused nor __block.  This change is to
   support compiling programs that are derived from BSD sources and use
diff --git a/ports/ChangeLog.aarch64 b/ports/ChangeLog.aarch64
index b1f6729..279a227 100644
--- a/ports/ChangeLog.aarch64
+++ b/ports/ChangeLog.aarch64
@@ -1,3 +1,9 @@
+2013-12-18  Marcus Shawcroft  <marcus.shawcroft@linaro.org>
+
+	[BZ #15128]
+	* sysdeps/aarch64/dl-trampoline.S (_dl_runtime_resolve): Save and
+	restore q0-q7.
+
 2013-12-17  Marcus Shawcroft  <marcus.shawcroft@linaro.org>
 
 	* sysdeps/unix/sysv/linux/aarch64/nptl/lowlevellock.h
diff --git a/ports/sysdeps/aarch64/dl-trampoline.S b/ports/sysdeps/aarch64/dl-trampoline.S
index 94c69a0..923ca76 100644
--- a/ports/sysdeps/aarch64/dl-trampoline.S
+++ b/ports/sysdeps/aarch64/dl-trampoline.S
@@ -41,7 +41,7 @@ _dl_runtime_resolve:
 	cfi_rel_offset (lr, 8)
 
 	/* Save arguments.  */
-	stp	x8, x9, [sp, #-80]!
+	stp	x8, x9, [sp, #-(80+8*16)]!
 	cfi_adjust_cfa_offset (80)
 	cfi_rel_offset (x8, 0)
 	cfi_rel_offset (x9, 8)
@@ -62,11 +62,27 @@ _dl_runtime_resolve:
 	cfi_rel_offset (x0, 64)
 	cfi_rel_offset (x1, 72)
 
+	stp	q0, q1, [sp, #(80+0*16)]
+	cfi_rel_offset (q0, 80+0*16)
+	cfi_rel_offset (q1, 80+1*16)
+
+	stp	q2, q3, [sp, #(80+2*16)]
+	cfi_rel_offset (q0, 80+2*16)
+	cfi_rel_offset (q1, 80+3*16)
+
+	stp	q4, q5, [sp, #(80+4*16)]
+	cfi_rel_offset (q0, 80+4*16)
+	cfi_rel_offset (q1, 80+5*16)
+
+	stp	q6, q7, [sp, #(80+6*16)]
+	cfi_rel_offset (q0, 80+6*16)
+	cfi_rel_offset (q1, 80+7*16)
+
 	/* Get pointer to linker struct.  */
 	ldr	x0, [ip0, #-8]
 
 	/* Prepare to call _dl_fixup().  */
-	ldr	x1, [sp, 80]	/* Recover &PLTGOT[n] */
+	ldr	x1, [sp, 80+8*16]	/* Recover &PLTGOT[n] */
 
 	sub     x1, x1, ip0
 	add     x1, x1, x1, lsl #1
@@ -81,12 +97,16 @@ _dl_runtime_resolve:
 	mov	ip0, x0
 
 	/* Get arguments and return address back.  */
+	ldp	q0, q1, [sp, #(80+0*16)]
+	ldp	q2, q3, [sp, #(80+2*16)]
+	ldp	q4, q5, [sp, #(80+4*16)]
+	ldp	q6, q7, [sp, #(80+6*16)]
 	ldp	x0, x1, [sp, #64]
 	ldp	x2, x3, [sp, #48]
 	ldp	x4, x5, [sp, #32]
 	ldp	x6, x7, [sp, #16]
-	ldp	x8, x9, [sp], #80
-	cfi_adjust_cfa_offset (-80)
+	ldp	x8, x9, [sp], #(80+8*16)
+	cfi_adjust_cfa_offset (-(80+8*16))
 
 	ldp	ip1, lr, [sp], #16
 	cfi_adjust_cfa_offset (-16)