Re: PING^1 [PATCH] testsuite: 'b' instruction can't do long enough jumps
On 2022-10-05 11:51, Kyrylo Tkachov wrote: Hi Torbjörn, -Original Message- From: Torbjorn SVENSSON Sent: Wednesday, October 5, 2022 10:28 AM To: Kyrylo Tkachov Cc: Yvan Roux Subject: Fwd: PING^1 [PATCH] testsuite: 'b' instruction can't do long enough jumps Hi Kyrill, I checked with Richard Sandiford if he could review this patch, but he pointed to you. Do you think that you can take a look it? This is ok. I don't think it changes any of the things we want to actually test for in these cases. Thanks, Kyrill Thank you! Pushed to master branch 1a46a0a8b30405ea353a758971634dabeee89eaf. Kind regards, Torbjörn
RE: PING^1 [PATCH] testsuite: 'b' instruction can't do long enough jumps
Hi Torbjörn, > -Original Message- > From: Torbjorn SVENSSON > Sent: Wednesday, October 5, 2022 10:28 AM > To: Kyrylo Tkachov > Cc: Yvan Roux > Subject: Fwd: PING^1 [PATCH] testsuite: 'b' instruction can't do long enough > jumps > > Hi Kyrill, > > I checked with Richard Sandiford if he could review this patch, but he > pointed to you. > Do you think that you can take a look it? This is ok. I don't think it changes any of the things we want to actually test for in these cases. Thanks, Kyrill > > Thank you! > > Kind regards, > Torbjörn > > > Forwarded Message > Subject: PING^1 [PATCH] testsuite: 'b' instruction can't do long enough > jumps > Date: Wed, 28 Sep 2022 11:17:13 +0200 > From: Torbjorn SVENSSON > To: gcc-patches@gcc.gnu.org > CC: yvan.r...@foss.st.com, r...@cebitec.uni-bielefeld.de, > mikest...@comcast.net, kyrylo.tkac...@arm.com > > Hi, > > Ping: https://gcc.gnu.org/pipermail/gcc-patches/2022- > September/601829.html > > Kind regards, > Torbjörn > > On 2022-09-19 18:30, Torbjörn SVENSSON wrote: > > After moving the testglue in commit 9d503515cee, the jump to exit and > > abort is too far for the 'b' instruction on Cortex-M0. As most of the > > C code would generate a 'bl' instruction instead of a 'b' > > instruction, lets do the same for the inline assembler. > > > > The error seen without this patch: > > > > /tmp/cccCRiCl.o: in function `main': > > stack-protector-1.c:(.text+0x4e): relocation truncated to fit: > R_ARM_THM_JUMP11 against symbol `__wrap_exit' defined in .text section > in gcc_tg.o > > stack-protector-1.c:(.text+0x50): relocation truncated to fit: > R_ARM_THM_JUMP11 against symbol `__wrap_abort' defined in .text > section in gcc_tg.o > > collect2: error: ld returned 1 exit status > > > > gcc/testsuite/ChangeLog: > > > > * gcc/testsuite/gcc.target/arm/stack-protector-1.c: Use 'bl' > > instead of 'b' instruction. > > * gcc/testsuite/gcc.target/arm/stack-protector-3.c: Likewise. > > > > Co-Authored-By: Yvan ROUX > > Signed-off-by: Torbjörn SVENSSON > > --- > > gcc/testsuite/gcc.target/arm/stack-protector-1.c | 4 ++-- > > gcc/testsuite/gcc.target/arm/stack-protector-3.c | 2 +- > > 2 files changed, 3 insertions(+), 3 deletions(-) > > > > diff --git a/gcc/testsuite/gcc.target/arm/stack-protector-1.c > b/gcc/testsuite/gcc.target/arm/stack-protector-1.c > > index 8d28b0a847c..3f0ffc9c3f3 100644 > > --- a/gcc/testsuite/gcc.target/arm/stack-protector-1.c > > +++ b/gcc/testsuite/gcc.target/arm/stack-protector-1.c > > @@ -56,8 +56,8 @@ asm ( > > " ldr r1, [sp, #4]\n" > > CHECK (r1) > > " mov r0, #0\n" > > -" b exit\n" > > +" bl exit\n" > > "1:\n" > > -" b abort\n" > > +" bl abort\n" > > " .size main, .-main" > > ); > > diff --git a/gcc/testsuite/gcc.target/arm/stack-protector-3.c > b/gcc/testsuite/gcc.target/arm/stack-protector-3.c > > index b8f77fa2309..2f710529b8f 100644 > > --- a/gcc/testsuite/gcc.target/arm/stack-protector-3.c > > +++ b/gcc/testsuite/gcc.target/arm/stack-protector-3.c > > @@ -26,7 +26,7 @@ asm ( > > " .type __stack_chk_fail, %function\n" > > "__stack_chk_fail:\n" > > " movsr0, #0\n" > > -" b exit\n" > > +" bl exit\n" > > " .size __stack_chk_fail, .-__stack_chk_fail" > > ); > >
Re: PING^1 [PATCH] testsuite: 'b' instruction can't do long enough jumps
On 9/28/22 15:39, Torbjorn SVENSSON wrote: Hi Christophe! On 2022-09-28 13:55, Christophe Lyon wrote: Hi! On 9/28/22 11:17, Torbjorn SVENSSON via Gcc-patches wrote: Hi, Ping: https://gcc.gnu.org/pipermail/gcc-patches/2022-September/601829.html Kind regards, Torbjörn On 2022-09-19 18:30, Torbjörn SVENSSON wrote: After moving the testglue in commit 9d503515cee, the jump to exit and abort is too far for the 'b' instruction on Cortex-M0. As most of the I am not sure I understand why that commit changed the distance between 'exit' and the branch instruction? The change was that the gcc_tg.o (the DejaGNU testglue.c object file) is now put last on the command line. In the previous versions of GCC, it was put before the ldflags flag etc, so now the code is placed in a way that the distance might be too big. This could also be related to that we in ST are using QEMU in system mode and not user mode and as a result, our test environment is slightly larger and might perhaps be placed in between the code for the test case and the testglue. Thanks, that makes sense. C code would generate a 'bl' instruction instead of a 'b' instruction, lets do the same for the inline assembler. The error seen without this patch: /tmp/cccCRiCl.o: in function `main': stack-protector-1.c:(.text+0x4e): relocation truncated to fit: R_ARM_THM_JUMP11 against symbol `__wrap_exit' defined in .text section in gcc_tg.o stack-protector-1.c:(.text+0x50): relocation truncated to fit: R_ARM_THM_JUMP11 against symbol `__wrap_abort' defined in .text section in gcc_tg.o collect2: error: ld returned 1 exit status Anyway the change seems sensible to me, I suppose it's not worth adding support in the linker to insert long branch stubs for these relocations. If a simple 'bl' instead of 'b' is enough, I think that this trivial change is the right one as the test case is supposed to test the stack protection, not branching, right? Yeah, agreed, I just meant to say a linker patch in addition to this one is probably not worth the effort. Christophe Kind regards, Torbjörn Christophe gcc/testsuite/ChangeLog: * gcc/testsuite/gcc.target/arm/stack-protector-1.c: Use 'bl' instead of 'b' instruction. * gcc/testsuite/gcc.target/arm/stack-protector-3.c: Likewise. Co-Authored-By: Yvan ROUX Signed-off-by: Torbjörn SVENSSON --- gcc/testsuite/gcc.target/arm/stack-protector-1.c | 4 ++-- gcc/testsuite/gcc.target/arm/stack-protector-3.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/gcc/testsuite/gcc.target/arm/stack-protector-1.c b/gcc/testsuite/gcc.target/arm/stack-protector-1.c index 8d28b0a847c..3f0ffc9c3f3 100644 --- a/gcc/testsuite/gcc.target/arm/stack-protector-1.c +++ b/gcc/testsuite/gcc.target/arm/stack-protector-1.c @@ -56,8 +56,8 @@ asm ( " ldr r1, [sp, #4]\n" CHECK (r1) " mov r0, #0\n" -" b exit\n" +" bl exit\n" "1:\n" -" b abort\n" +" bl abort\n" " .size main, .-main" ); diff --git a/gcc/testsuite/gcc.target/arm/stack-protector-3.c b/gcc/testsuite/gcc.target/arm/stack-protector-3.c index b8f77fa2309..2f710529b8f 100644 --- a/gcc/testsuite/gcc.target/arm/stack-protector-3.c +++ b/gcc/testsuite/gcc.target/arm/stack-protector-3.c @@ -26,7 +26,7 @@ asm ( " .type __stack_chk_fail, %function\n" "__stack_chk_fail:\n" " movs r0, #0\n" -" b exit\n" +" bl exit\n" " .size __stack_chk_fail, .-__stack_chk_fail" );
Re: PING^1 [PATCH] testsuite: 'b' instruction can't do long enough jumps
Hi Christophe! On 2022-09-28 13:55, Christophe Lyon wrote: Hi! On 9/28/22 11:17, Torbjorn SVENSSON via Gcc-patches wrote: Hi, Ping: https://gcc.gnu.org/pipermail/gcc-patches/2022-September/601829.html Kind regards, Torbjörn On 2022-09-19 18:30, Torbjörn SVENSSON wrote: After moving the testglue in commit 9d503515cee, the jump to exit and abort is too far for the 'b' instruction on Cortex-M0. As most of the I am not sure I understand why that commit changed the distance between 'exit' and the branch instruction? The change was that the gcc_tg.o (the DejaGNU testglue.c object file) is now put last on the command line. In the previous versions of GCC, it was put before the ldflags flag etc, so now the code is placed in a way that the distance might be too big. This could also be related to that we in ST are using QEMU in system mode and not user mode and as a result, our test environment is slightly larger and might perhaps be placed in between the code for the test case and the testglue. C code would generate a 'bl' instruction instead of a 'b' instruction, lets do the same for the inline assembler. The error seen without this patch: /tmp/cccCRiCl.o: in function `main': stack-protector-1.c:(.text+0x4e): relocation truncated to fit: R_ARM_THM_JUMP11 against symbol `__wrap_exit' defined in .text section in gcc_tg.o stack-protector-1.c:(.text+0x50): relocation truncated to fit: R_ARM_THM_JUMP11 against symbol `__wrap_abort' defined in .text section in gcc_tg.o collect2: error: ld returned 1 exit status Anyway the change seems sensible to me, I suppose it's not worth adding support in the linker to insert long branch stubs for these relocations. If a simple 'bl' instead of 'b' is enough, I think that this trivial change is the right one as the test case is supposed to test the stack protection, not branching, right? Kind regards, Torbjörn Christophe gcc/testsuite/ChangeLog: * gcc/testsuite/gcc.target/arm/stack-protector-1.c: Use 'bl' instead of 'b' instruction. * gcc/testsuite/gcc.target/arm/stack-protector-3.c: Likewise. Co-Authored-By: Yvan ROUX Signed-off-by: Torbjörn SVENSSON --- gcc/testsuite/gcc.target/arm/stack-protector-1.c | 4 ++-- gcc/testsuite/gcc.target/arm/stack-protector-3.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/gcc/testsuite/gcc.target/arm/stack-protector-1.c b/gcc/testsuite/gcc.target/arm/stack-protector-1.c index 8d28b0a847c..3f0ffc9c3f3 100644 --- a/gcc/testsuite/gcc.target/arm/stack-protector-1.c +++ b/gcc/testsuite/gcc.target/arm/stack-protector-1.c @@ -56,8 +56,8 @@ asm ( " ldr r1, [sp, #4]\n" CHECK (r1) " mov r0, #0\n" -" b exit\n" +" bl exit\n" "1:\n" -" b abort\n" +" bl abort\n" " .size main, .-main" ); diff --git a/gcc/testsuite/gcc.target/arm/stack-protector-3.c b/gcc/testsuite/gcc.target/arm/stack-protector-3.c index b8f77fa2309..2f710529b8f 100644 --- a/gcc/testsuite/gcc.target/arm/stack-protector-3.c +++ b/gcc/testsuite/gcc.target/arm/stack-protector-3.c @@ -26,7 +26,7 @@ asm ( " .type __stack_chk_fail, %function\n" "__stack_chk_fail:\n" " movs r0, #0\n" -" b exit\n" +" bl exit\n" " .size __stack_chk_fail, .-__stack_chk_fail" );
Re: PING^1 [PATCH] testsuite: 'b' instruction can't do long enough jumps
Hi! On 9/28/22 11:17, Torbjorn SVENSSON via Gcc-patches wrote: Hi, Ping: https://gcc.gnu.org/pipermail/gcc-patches/2022-September/601829.html Kind regards, Torbjörn On 2022-09-19 18:30, Torbjörn SVENSSON wrote: After moving the testglue in commit 9d503515cee, the jump to exit and abort is too far for the 'b' instruction on Cortex-M0. As most of the I am not sure I understand why that commit changed the distance between 'exit' and the branch instruction? C code would generate a 'bl' instruction instead of a 'b' instruction, lets do the same for the inline assembler. The error seen without this patch: /tmp/cccCRiCl.o: in function `main': stack-protector-1.c:(.text+0x4e): relocation truncated to fit: R_ARM_THM_JUMP11 against symbol `__wrap_exit' defined in .text section in gcc_tg.o stack-protector-1.c:(.text+0x50): relocation truncated to fit: R_ARM_THM_JUMP11 against symbol `__wrap_abort' defined in .text section in gcc_tg.o collect2: error: ld returned 1 exit status Anyway the change seems sensible to me, I suppose it's not worth adding support in the linker to insert long branch stubs for these relocations. Christophe gcc/testsuite/ChangeLog: * gcc/testsuite/gcc.target/arm/stack-protector-1.c: Use 'bl' instead of 'b' instruction. * gcc/testsuite/gcc.target/arm/stack-protector-3.c: Likewise. Co-Authored-By: Yvan ROUX Signed-off-by: Torbjörn SVENSSON --- gcc/testsuite/gcc.target/arm/stack-protector-1.c | 4 ++-- gcc/testsuite/gcc.target/arm/stack-protector-3.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/gcc/testsuite/gcc.target/arm/stack-protector-1.c b/gcc/testsuite/gcc.target/arm/stack-protector-1.c index 8d28b0a847c..3f0ffc9c3f3 100644 --- a/gcc/testsuite/gcc.target/arm/stack-protector-1.c +++ b/gcc/testsuite/gcc.target/arm/stack-protector-1.c @@ -56,8 +56,8 @@ asm ( " ldr r1, [sp, #4]\n" CHECK (r1) " mov r0, #0\n" -" b exit\n" +" bl exit\n" "1:\n" -" b abort\n" +" bl abort\n" " .size main, .-main" ); diff --git a/gcc/testsuite/gcc.target/arm/stack-protector-3.c b/gcc/testsuite/gcc.target/arm/stack-protector-3.c index b8f77fa2309..2f710529b8f 100644 --- a/gcc/testsuite/gcc.target/arm/stack-protector-3.c +++ b/gcc/testsuite/gcc.target/arm/stack-protector-3.c @@ -26,7 +26,7 @@ asm ( " .type __stack_chk_fail, %function\n" "__stack_chk_fail:\n" " movs r0, #0\n" -" b exit\n" +" bl exit\n" " .size __stack_chk_fail, .-__stack_chk_fail" );
