Skip to contentSkip to conversation

sim: or1k: Eliminate dangerous RWX load segments

PATCH In review Stafford Horne 3 participants 5 messages Aug 19, 2023 – Oct 14, 2023

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 sim: or1k: Eliminate dangerous RWX load segments

This fixes test failures caused by the new linker warning which report:

./ld/ld-new: warning: load.S.x has a LOAD segment with RWX permissions

Fix this by splitting the linker MEMORY into ram and rom to avoid generating RWX sections. This required tests to be adjusted to fix issues with the move. Namely:

- fpu tests: were incorrectly using l.ori with ha(anchor) which now that we pushed the anchor up in memory it exposes the bug. Update to used the correct l.movhi instruction instead.
- adrp test: the test reports ram offset addresses, now that we have moved memory layout around a bit I adjusted the test output. Some padding is added before pi to show that the actual address of pi and the adrp page offset are not the same.

---
 sim/testsuite/or1k/adrp.S               | 5 +++--
 sim/testsuite/or1k/fpu-unordered.S      | 2 +-
 sim/testsuite/or1k/fpu64a32-unordered.S | 2 +-
 sim/testsuite/or1k/fpu64a32.S           | 2 +-
 sim/testsuite/or1k/or1k-test.ld         | 7 ++++---
 5 files changed, 10 insertions(+), 8 deletions(-)
sim/testsuite/or1k/adrp.S +3 -2
index eaddcb03885..192324c698e 100644
@@ -17,9 +17,9 @@
 
 # mach: or1k
 # output: report(0x00002064);\n
-# output: report(0x00012138);\n
+# output: report(0x0001a008);\n
 # output: report(0x00002000);\n
-# output: report(0x00012000);\n
+# output: report(0x0001a000);\n
 # output: report(0x00002000);\n
 # output: report(0x00014000);\n
 # output: report(0x00000000);\n
@@ -32,6 +32,7 @@
 	.section .data
 	.org 0x10000
 	.align 4
+pad:	.quad 	0
 	.type   pi, @object
 	.size   pi, 4
 pi:
sim/testsuite/or1k/fpu-unordered.S +1 -1
index 624aa0fe05d..a89172e37af 100644
@@ -57,7 +57,7 @@ start_tests:
 	 *  r13  e   as float
 	 *  r16  nan as float
 	 */
-	l.ori	r11, r0, ha(anchor)
+	l.movhi	r11, ha(anchor)
 	l.addi	r11, r11, lo(anchor)
 	l.lwz	r12, 0(r11)
 
sim/testsuite/or1k/fpu64a32-unordered.S +1 -1
index e0ae6e770d1..51d915e4e75 100644
@@ -58,7 +58,7 @@ start_tests:
 	 *  r14,r15  e   as double
 	 *  r16,r17  nan as double
 	 */
-	l.ori	r11, r0, ha(anchor)
+	l.movhi	r11, ha(anchor)
 	l.addi	r11, r11, lo(anchor)
 	l.lwz	r12, 0(r11)
 	l.lwz	r13, 4(r11)
sim/testsuite/or1k/fpu64a32.S +1 -1
index 71b72b7761c..6ea60b28cf2 100644
@@ -98,7 +98,7 @@ start_tests:
 	 *  r14,r15  e  as double
 	 *  r16,r17  a long long
 	 */
-	l.ori	r11, r0, ha(anchor)
+	l.movhi	r11, ha(anchor)
 	l.addi	r11, r11, lo(anchor)
 	l.lwz	r12, 0(r11)
 	l.lwz	r13, 4(r11)
sim/testsuite/or1k/or1k-test.ld +4 -3
index f1535daeabd..c26ecaf3f23 100644
@@ -20,8 +20,9 @@ MEMORY
     /* The exception vectors actually start at 0x100, but if you specify
        that address here, the "--output-target binary" step will start from
        address 0 with the contents meant for address 0x100.  */
-    exception_vectors : ORIGIN =  0 , LENGTH = 8K
-    ram               : ORIGIN =  8K, LENGTH = 2M - 8K
+    exception_vectors : ORIGIN =  0 , LENGTH =  8K
+    rom               : ORIGIN =  8K, LENGTH = 40K
+    ram               : ORIGIN = 40K, LENGTH =  2M - 40K
 }
 
 SECTIONS
@@ -37,7 +38,7 @@ SECTIONS
 	*(.text.*)
 	*(.rodata)
 	*(.rodata.*)
-    } > ram
+    } > rom
 
     .data :
     {
Signature · 1 line
2.39.1
#1 lore raw

Hi Stafford Horne,

The simulator code is actually maintained by the GNU GDB project rather than the GNU Binutils project, so I am redirecting this email to the gdb-patches list.

Quoted 8 lines

This fixes test failures caused by the new linker warning which report:

./ld/ld-new: warning: load.S.x has a LOAD segment with RWX permissions

Fix this by splitting the linker MEMORY into ram and rom to avoid generating RWX sections. This required tests to be adjusted to fix issues with the move. Namely:

Quoted 7 lines

- fpu tests: were incorrectly using l.ori with ha(anchor) which now that we pushed the anchor up in memory it exposes the bug. Update to used the correct l.movhi instruction instead.
- adrp test: the test reports ram offset addresses, now that we have moved memory layout around a bit I adjusted the test output. Some padding is added before pi to show that the actual address of pi and the adrp page offset are not the same.

(It is nice to see that this new linker feature helped to detect these problems. :-)

Quoted 97 lines
---
  sim/testsuite/or1k/adrp.S               | 5 +++--
  sim/testsuite/or1k/fpu-unordered.S      | 2 +-
  sim/testsuite/or1k/fpu64a32-unordered.S | 2 +-
  sim/testsuite/or1k/fpu64a32.S           | 2 +-
  sim/testsuite/or1k/or1k-test.ld         | 7 ++++---
  5 files changed, 10 insertions(+), 8 deletions(-)
sim/testsuite/or1k/adrp.S +3 -2
index eaddcb03885..192324c698e 100644
@@ -17,9 +17,9 @@
 
  # mach: or1k
  # output: report(0x00002064);\n
-# output: report(0x00012138);\n
+# output: report(0x0001a008);\n
  # output: report(0x00002000);\n
-# output: report(0x00012000);\n
+# output: report(0x0001a000);\n
  # output: report(0x00002000);\n
  # output: report(0x00014000);\n
  # output: report(0x00000000);\n
@@ -32,6 +32,7 @@
  	.section .data
  	.org 0x10000
  	.align 4
+pad:	.quad 	0
  	.type   pi, @object
  	.size   pi, 4
  pi:
sim/testsuite/or1k/fpu-unordered.S +1 -1
index 624aa0fe05d..a89172e37af 100644
@@ -57,7 +57,7 @@ start_tests:
  	 *  r13  e   as float
  	 *  r16  nan as float
  	 */
-	l.ori	r11, r0, ha(anchor)
+	l.movhi	r11, ha(anchor)
  	l.addi	r11, r11, lo(anchor)
  	l.lwz	r12, 0(r11)
 
sim/testsuite/or1k/fpu64a32-unordered.S +1 -1
index e0ae6e770d1..51d915e4e75 100644
@@ -58,7 +58,7 @@ start_tests:
  	 *  r14,r15  e   as double
  	 *  r16,r17  nan as double
  	 */
-	l.ori	r11, r0, ha(anchor)
+	l.movhi	r11, ha(anchor)
  	l.addi	r11, r11, lo(anchor)
  	l.lwz	r12, 0(r11)
  	l.lwz	r13, 4(r11)
sim/testsuite/or1k/fpu64a32.S +1 -1
index 71b72b7761c..6ea60b28cf2 100644
@@ -98,7 +98,7 @@ start_tests:
  	 *  r14,r15  e  as double
  	 *  r16,r17  a long long
  	 */
-	l.ori	r11, r0, ha(anchor)
+	l.movhi	r11, ha(anchor)
  	l.addi	r11, r11, lo(anchor)
  	l.lwz	r12, 0(r11)
  	l.lwz	r13, 4(r11)
sim/testsuite/or1k/or1k-test.ld +4 -3
index f1535daeabd..c26ecaf3f23 100644
@@ -20,8 +20,9 @@ MEMORY
      /* The exception vectors actually start at 0x100, but if you specify
         that address here, the "--output-target binary" step will start from
         address 0 with the contents meant for address 0x100.  */
-    exception_vectors : ORIGIN =  0 , LENGTH = 8K
-    ram               : ORIGIN =  8K, LENGTH = 2M - 8K
+    exception_vectors : ORIGIN =  0 , LENGTH =  8K
+    rom               : ORIGIN =  8K, LENGTH = 40K
+    ram               : ORIGIN = 40K, LENGTH =  2M - 40K
  }
 
  SECTIONS
@@ -37,7 +38,7 @@ SECTIONS
  	*(.text.*)
  	*(.rodata)
  	*(.rodata.*)
-    } > ram
+    } > rom
 
      .data :
      {
#2 lore raw
In reply to Stafford Horne
Quoted 16 lines · Stafford Horne

This fixes test failures caused by the new linker warning which report:

./ld/ld-new: warning: load.S.x has a LOAD segment with RWX permissions

Fix this by splitting the linker MEMORY into ram and rom to avoid generating RWX sections. This required tests to be adjusted to fix issues with the move. Namely:

- fpu tests: were incorrectly using l.ori with ha(anchor) which now that we pushed the anchor up in memory it exposes the bug. Update to used the correct l.movhi instruction instead.
- adrp test: the test reports ram offset addresses, now that we have moved memory layout around a bit I adjusted the test output. Some padding is added before pi to show that the actual address of pi and the adrp page offset are not the same.

Quoted 1 line

nit: put [PR sim/29957] at the end of the git commit summary (first line)

otherwise, lgtm, thanks
-mike

#3 lore raw
Quoted 17 lines · Mike Frysinger
Quoted 16 lines · Stafford Horne

This fixes test failures caused by the new linker warning which report:

./ld/ld-new: warning: load.S.x has a LOAD segment with RWX permissions

Fix this by splitting the linker MEMORY into ram and rom to avoid generating RWX sections. This required tests to be adjusted to fix issues with the move. Namely:

- fpu tests: were incorrectly using l.ori with ha(anchor) which now that we pushed the anchor up in memory it exposes the bug. Update to used the correct l.movhi instruction instead.
- adrp test: the test reports ram offset addresses, now that we have moved memory layout around a bit I adjusted the test output. Some padding is added before pi to show that the actual address of pi and the adrp page offset are not the same.

Quoted 6 lines
Quoted 1 line

nit: put [PR sim/29957] at the end of the git commit summary (first line)

otherwise, lgtm, thanks
-mike

Thanks Mike,

I actually pushed this long back without the above update after I read Nick's comment. I for some reason missed that he said he was forwarding to the sim maintainers.

Sorry for that.

-Stafford

#4 lore raw
Quoted 25 lines · Stafford Horne
Quoted 17 lines · Mike Frysinger
Quoted 16 lines · Stafford Horne

This fixes test failures caused by the new linker warning which report:

./ld/ld-new: warning: load.S.x has a LOAD segment with RWX permissions

Fix this by splitting the linker MEMORY into ram and rom to avoid generating RWX sections. This required tests to be adjusted to fix issues with the move. Namely:

- fpu tests: were incorrectly using l.ori with ha(anchor) which now that we pushed the anchor up in memory it exposes the bug. Update to used the correct l.movhi instruction instead.
- adrp test: the test reports ram offset addresses, now that we have moved memory layout around a bit I adjusted the test output. Some padding is added before pi to show that the actual address of pi and the adrp page offset are not the same.

Quoted 6 lines
Quoted 1 line

nit: put [PR sim/29957] at the end of the git commit summary (first line)

otherwise, lgtm, thanks
-mike

Quoted 7 lines

Thanks Mike,

I actually pushed this long back without the above update after I read Nick's comment. I for some reason missed that he said he was forwarding to the sim maintainers.

Sorry for that.

np. had some hardware failures on my main desktop so only just catching up. -mike