On x86-64, the parametrized selftest code for rseq crashes with a segmentation fault when compiled with -fpie. This happens when the param_test binary is loaded at an address beyond 32-bit on x86-64.
The issue is caused by use of a 32-bit register to hold the address of the loop counter variable.
Fix this by using a 64-bit register to calculate the address of the loop counter variables as an offset from rip.
Signed-off-by: Mathieu Desnoyers mathieu.desnoyers@efficios.com Cc: stable@vger.kernel.org # v4.18 Cc: Thomas Gleixner tglx@linutronix.de Cc: Joel Fernandes joelaf@google.com Cc: Peter Zijlstra peterz@infradead.org Cc: Catalin Marinas catalin.marinas@arm.com Cc: Dave Watson davejwatson@fb.com Cc: Will Deacon will.deacon@arm.com Cc: Shuah Khan shuahkh@osg.samsung.com Cc: Andi Kleen andi@firstfloor.org Cc: linux-kselftest@vger.kernel.org Cc: "H . Peter Anvin" hpa@zytor.com Cc: Chris Lameter cl@linux.com Cc: Russell King linux@arm.linux.org.uk Cc: Michael Kerrisk mtk.manpages@gmail.com Cc: "Paul E . McKenney" paulmck@linux.vnet.ibm.com Cc: Paul Turner pjt@google.com Cc: Boqun Feng boqun.feng@gmail.com Cc: Josh Triplett josh@joshtriplett.org Cc: Steven Rostedt rostedt@goodmis.org Cc: Ben Maurer bmaurer@fb.com Cc: Andy Lutomirski luto@amacapital.net Cc: Andrew Morton akpm@linux-foundation.org Cc: Linus Torvalds torvalds@linux-foundation.org --- tools/testing/selftests/rseq/param_test.c | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-)
diff --git a/tools/testing/selftests/rseq/param_test.c b/tools/testing/selftests/rseq/param_test.c index 615252331813..4bc071525bf7 100644 --- a/tools/testing/selftests/rseq/param_test.c +++ b/tools/testing/selftests/rseq/param_test.c @@ -56,15 +56,13 @@ unsigned int yield_mod_cnt, nr_abort; printf(fmt, ## __VA_ARGS__); \ } while (0)
-#if defined(__x86_64__) || defined(__i386__) +#ifdef __i386__
#define INJECT_ASM_REG "eax"
#define RSEQ_INJECT_CLOBBER \ , INJECT_ASM_REG
-#ifdef __i386__ - #define RSEQ_INJECT_ASM(n) \ "mov asm_loop_cnt_" #n ", %%" INJECT_ASM_REG "\n\t" \ "test %%" INJECT_ASM_REG ",%%" INJECT_ASM_REG "\n\t" \ @@ -76,9 +74,16 @@ unsigned int yield_mod_cnt, nr_abort;
#elif defined(__x86_64__)
+#define INJECT_ASM_REG_P "rax" +#define INJECT_ASM_REG "eax" + +#define RSEQ_INJECT_CLOBBER \ + , INJECT_ASM_REG_P \ + , INJECT_ASM_REG + #define RSEQ_INJECT_ASM(n) \ - "lea asm_loop_cnt_" #n "(%%rip), %%" INJECT_ASM_REG "\n\t" \ - "mov (%%" INJECT_ASM_REG "), %%" INJECT_ASM_REG "\n\t" \ + "lea asm_loop_cnt_" #n "(%%rip), %%" INJECT_ASM_REG_P "\n\t" \ + "mov (%%" INJECT_ASM_REG_P "), %%" INJECT_ASM_REG "\n\t" \ "test %%" INJECT_ASM_REG ",%%" INJECT_ASM_REG "\n\t" \ "jz 333f\n\t" \ "222:\n\t" \ @@ -86,10 +91,6 @@ unsigned int yield_mod_cnt, nr_abort; "jnz 222b\n\t" \ "333:\n\t"
-#else -#error "Unsupported architecture" -#endif - #elif defined(__ARMEL__)
#define RSEQ_INJECT_INPUT \
----- On Sep 18, 2018, at 9:53 AM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
On x86-64, the parametrized selftest code for rseq crashes with a segmentation fault when compiled with -fpie. This happens when the param_test binary is loaded at an address beyond 32-bit on x86-64.
The issue is caused by use of a 32-bit register to hold the address of the loop counter variable.
Fix this by using a 64-bit register to calculate the address of the loop counter variables as an offset from rip.
Should this fix go through tip or the selftests tree ?
Thanks,
Mathieu
Signed-off-by: Mathieu Desnoyers mathieu.desnoyers@efficios.com Cc: stable@vger.kernel.org # v4.18 Cc: Thomas Gleixner tglx@linutronix.de Cc: Joel Fernandes joelaf@google.com Cc: Peter Zijlstra peterz@infradead.org Cc: Catalin Marinas catalin.marinas@arm.com Cc: Dave Watson davejwatson@fb.com Cc: Will Deacon will.deacon@arm.com Cc: Shuah Khan shuahkh@osg.samsung.com Cc: Andi Kleen andi@firstfloor.org Cc: linux-kselftest@vger.kernel.org Cc: "H . Peter Anvin" hpa@zytor.com Cc: Chris Lameter cl@linux.com Cc: Russell King linux@arm.linux.org.uk Cc: Michael Kerrisk mtk.manpages@gmail.com Cc: "Paul E . McKenney" paulmck@linux.vnet.ibm.com Cc: Paul Turner pjt@google.com Cc: Boqun Feng boqun.feng@gmail.com Cc: Josh Triplett josh@joshtriplett.org Cc: Steven Rostedt rostedt@goodmis.org Cc: Ben Maurer bmaurer@fb.com Cc: Andy Lutomirski luto@amacapital.net Cc: Andrew Morton akpm@linux-foundation.org Cc: Linus Torvalds torvalds@linux-foundation.org
tools/testing/selftests/rseq/param_test.c | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-)
diff --git a/tools/testing/selftests/rseq/param_test.c b/tools/testing/selftests/rseq/param_test.c index 615252331813..4bc071525bf7 100644 --- a/tools/testing/selftests/rseq/param_test.c +++ b/tools/testing/selftests/rseq/param_test.c @@ -56,15 +56,13 @@ unsigned int yield_mod_cnt, nr_abort; printf(fmt, ## __VA_ARGS__); \ } while (0)
-#if defined(__x86_64__) || defined(__i386__) +#ifdef __i386__
#define INJECT_ASM_REG "eax"
#define RSEQ_INJECT_CLOBBER \ , INJECT_ASM_REG
-#ifdef __i386__
#define RSEQ_INJECT_ASM(n) \ "mov asm_loop_cnt_" #n ", %%" INJECT_ASM_REG "\n\t" \ "test %%" INJECT_ASM_REG ",%%" INJECT_ASM_REG "\n\t" \ @@ -76,9 +74,16 @@ unsigned int yield_mod_cnt, nr_abort;
#elif defined(__x86_64__)
+#define INJECT_ASM_REG_P "rax" +#define INJECT_ASM_REG "eax"
+#define RSEQ_INJECT_CLOBBER \
- , INJECT_ASM_REG_P \
- , INJECT_ASM_REG
#define RSEQ_INJECT_ASM(n) \
- "lea asm_loop_cnt_" #n "(%%rip), %%" INJECT_ASM_REG "\n\t" \
- "mov (%%" INJECT_ASM_REG "), %%" INJECT_ASM_REG "\n\t" \
- "lea asm_loop_cnt_" #n "(%%rip), %%" INJECT_ASM_REG_P "\n\t" \
- "mov (%%" INJECT_ASM_REG_P "), %%" INJECT_ASM_REG "\n\t" \ "test %%" INJECT_ASM_REG ",%%" INJECT_ASM_REG "\n\t" \ "jz 333f\n\t" \ "222:\n\t" \
@@ -86,10 +91,6 @@ unsigned int yield_mod_cnt, nr_abort; "jnz 222b\n\t" \ "333:\n\t"
-#else -#error "Unsupported architecture" -#endif
#elif defined(__ARMEL__)
#define RSEQ_INJECT_INPUT \
2.11.0
On Tue, 25 Sep 2018 13:39:36 -0400 (EDT) Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
----- On Sep 18, 2018, at 9:53 AM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
On x86-64, the parametrized selftest code for rseq crashes with a segmentation fault when compiled with -fpie. This happens when the param_test binary is loaded at an address beyond 32-bit on x86-64.
The issue is caused by use of a 32-bit register to hold the address of the loop counter variable.
Fix this by using a 64-bit register to calculate the address of the loop counter variables as an offset from rip.
Should this fix go through tip or the selftests tree ?
I usually have changes like this pulled through Shuah's tree with an ack from the maintainer of the code that it tests.
-- Steve
On 09/27/2018 07:22 AM, Steven Rostedt wrote:
On Tue, 25 Sep 2018 13:39:36 -0400 (EDT) Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
----- On Sep 18, 2018, at 9:53 AM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
On x86-64, the parametrized selftest code for rseq crashes with a segmentation fault when compiled with -fpie. This happens when the param_test binary is loaded at an address beyond 32-bit on x86-64.
The issue is caused by use of a 32-bit register to hold the address of the loop counter variable.
Fix this by using a 64-bit register to calculate the address of the loop counter variables as an offset from rip.
Should this fix go through tip or the selftests tree ?
I usually have changes like this pulled through Shuah's tree with an ack from the maintainer of the code that it tests.
-- Steve
Right that is what I prefer. I usually wait for an Ack. I can pull this in.
thanks, -- Shuah
----- On Sep 27, 2018, at 9:58 AM, Shuah Khan shuah.kh@samsung.com wrote:
On 09/27/2018 07:22 AM, Steven Rostedt wrote:
On Tue, 25 Sep 2018 13:39:36 -0400 (EDT) Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
----- On Sep 18, 2018, at 9:53 AM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
On x86-64, the parametrized selftest code for rseq crashes with a segmentation fault when compiled with -fpie. This happens when the param_test binary is loaded at an address beyond 32-bit on x86-64.
The issue is caused by use of a 32-bit register to hold the address of the loop counter variable.
Fix this by using a 64-bit register to calculate the address of the loop counter variables as an offset from rip.
Should this fix go through tip or the selftests tree ?
I usually have changes like this pulled through Shuah's tree with an ack from the maintainer of the code that it tests.
-- Steve
Right that is what I prefer. I usually wait for an Ack. I can pull this in.
I am maintainer of the code that it tests as well, so being both author of the selftest patch and giving an ack to it seems a bit silly. :) Other rseq co-maintainers could weight in though: Peter, Paul, and Boqun.
Thanks,
Mathieu
thanks, -- Shuah
-- Shuah Khan Sr. Linux Kernel Developer Open Source Innovation Group Samsung Research America(Silicon Valley) shuah.kh@samsung.com
On 09/27/2018 12:02 PM, Mathieu Desnoyers wrote:
----- On Sep 27, 2018, at 9:58 AM, Shuah Khan shuah.kh@samsung.com wrote:
On 09/27/2018 07:22 AM, Steven Rostedt wrote:
On Tue, 25 Sep 2018 13:39:36 -0400 (EDT) Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
----- On Sep 18, 2018, at 9:53 AM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
On x86-64, the parametrized selftest code for rseq crashes with a segmentation fault when compiled with -fpie. This happens when the param_test binary is loaded at an address beyond 32-bit on x86-64.
The issue is caused by use of a 32-bit register to hold the address of the loop counter variable.
Fix this by using a 64-bit register to calculate the address of the loop counter variables as an offset from rip.
Should this fix go through tip or the selftests tree ?
I usually have changes like this pulled through Shuah's tree with an ack from the maintainer of the code that it tests.
-- Steve
Right that is what I prefer. I usually wait for an Ack. I can pull this in.
I am maintainer of the code that it tests as well, so being both author of the selftest patch and giving an ack to it seems a bit silly. :) Other rseq co-maintainers could weight in though: Peter, Paul, and Boqun.
I was commenting in general, not specific to this patch. No worries. I will apply this right away.
I have been traveling the last couple of weeks and in the catch up mode still.
thanks, -- Shuah
-- Shuah Khan Sr. Linux Kernel Developer Open Source Innovation Group Samsung Research America(Silicon Valley) shuah.kh@samsung.com
On 09/27/2018 12:10 PM, Shuah Khan wrote:
On 09/27/2018 12:02 PM, Mathieu Desnoyers wrote:
----- On Sep 27, 2018, at 9:58 AM, Shuah Khan shuah.kh@samsung.com wrote:
On 09/27/2018 07:22 AM, Steven Rostedt wrote:
On Tue, 25 Sep 2018 13:39:36 -0400 (EDT) Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
----- On Sep 18, 2018, at 9:53 AM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
On x86-64, the parametrized selftest code for rseq crashes with a segmentation fault when compiled with -fpie. This happens when the param_test binary is loaded at an address beyond 32-bit on x86-64.
The issue is caused by use of a 32-bit register to hold the address of the loop counter variable.
Fix this by using a 64-bit register to calculate the address of the loop counter variables as an offset from rip.
Should this fix go through tip or the selftests tree ?
I usually have changes like this pulled through Shuah's tree with an ack from the maintainer of the code that it tests.
-- Steve
Right that is what I prefer. I usually wait for an Ack. I can pull this in.
I am maintainer of the code that it tests as well, so being both author of the selftest patch and giving an ack to it seems a bit silly. :) Other rseq co-maintainers could weight in though: Peter, Paul, and Boqun.
I was commenting in general, not specific to this patch. No worries. I will apply this right away.
I have been traveling the last couple of weeks and in the catch up mode still.
Hmm. this patch doesn't apply to linux-kselftest fixes branch which is at 4.19-rc4. Also you don't have my current email from getmaintainers.
Could you please rebase and send it to me.
thanks, -- Shuah
On Thu, Sep 27, 2018 at 02:02:08PM -0400, Mathieu Desnoyers wrote:
----- On Sep 27, 2018, at 9:58 AM, Shuah Khan shuah.kh@samsung.com wrote:
On 09/27/2018 07:22 AM, Steven Rostedt wrote:
On Tue, 25 Sep 2018 13:39:36 -0400 (EDT) Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
----- On Sep 18, 2018, at 9:53 AM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
On x86-64, the parametrized selftest code for rseq crashes with a segmentation fault when compiled with -fpie. This happens when the param_test binary is loaded at an address beyond 32-bit on x86-64.
The issue is caused by use of a 32-bit register to hold the address of the loop counter variable.
Fix this by using a 64-bit register to calculate the address of the loop counter variables as an offset from rip.
Should this fix go through tip or the selftests tree ?
I usually have changes like this pulled through Shuah's tree with an ack from the maintainer of the code that it tests.
-- Steve
Right that is what I prefer. I usually wait for an Ack. I can pull this in.
I am maintainer of the code that it tests as well, so being both author of the selftest patch and giving an ack to it seems a bit silly. :) Other rseq co-maintainers could weight in though: Peter, Paul, and Boqun.
Here you go! ;-)
Acked-by: Paul E. McKenney paulmck@linux.ibm.com
Thanks,
Mathieu
thanks, -- Shuah
-- Shuah Khan Sr. Linux Kernel Developer Open Source Innovation Group Samsung Research America(Silicon Valley) shuah.kh@samsung.com
-- Mathieu Desnoyers EfficiOS Inc. http://www.efficios.com
linux-kselftest-mirror@lists.linaro.org
-
Mathieu Desnoyers -
Paul E. McKenney -
Shuah Khan -
Shuah Khan -
Steven Rostedt