Re: PING: [PATCH 2/2] tests/tcg/s390x: Test unwinding from signal handlers

[Thomas Huth <thuth@redhat.com>]

On 19/05/2022 13.34, Ilya Leoshkevich wrote:
> On Wed, 2022-05-04 at 00:51 +0200, Ilya Leoshkevich wrote:
>> Add a small test to prevent regressions.
>>
>> Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
>> ---
>>   tests/tcg/s390x/signals-s390x.c | 69 ++++++++++++++++++++++++++-----
>> --
>>   1 file changed, 55 insertions(+), 14 deletions(-)
>>
>> diff --git a/tests/tcg/s390x/signals-s390x.c
>> b/tests/tcg/s390x/signals-s390x.c
>> index dc2f8ee59a..48c3b6cdfd 100644
>> --- a/tests/tcg/s390x/signals-s390x.c
>> +++ b/tests/tcg/s390x/signals-s390x.c
>> @@ -1,4 +1,5 @@
>>   #include <assert.h>
>> +#include <execinfo.h>
>>   #include <signal.h>
>>   #include <string.h>
>>   #include <sys/mman.h>
>> @@ -11,22 +12,28 @@
>>    * inline asm is used instead.
>>    */
>>   
>> +#define DEFINE_ASM_FUNCTION(name, body) \
>> +    asm(".globl " #name "\n" \
>> +        #name ":\n" \
>> +        ".cfi_startproc\n" \
>> +        body "\n" \
>> +        "br %r14\n" \
>> +        ".cfi_endproc");
>> +
>>   void illegal_op(void);
>> -void after_illegal_op(void);
>> -asm(".globl\tillegal_op\n"
>> -    "illegal_op:\t.byte\t0x00,0x00\n"
>> -    "\t.globl\tafter_illegal_op\n"
>> -    "after_illegal_op:\tbr\t%r14");
>> +extern const char after_illegal_op;
>> +DEFINE_ASM_FUNCTION(illegal_op,
>> +    ".byte 0x00,0x00\n"
>> +    ".globl after_illegal_op\n"
>> +    "after_illegal_op:")
>>   
>>   void stg(void *dst, unsigned long src);
>> -asm(".globl\tstg\n"
>> -    "stg:\tstg\t%r3,0(%r2)\n"
>> -    "\tbr\t%r14");
>> +DEFINE_ASM_FUNCTION(stg, "stg %r3,0(%r2)")
>>   
>>   void mvc_8(void *dst, void *src);
>> -asm(".globl\tmvc_8\n"
>> -    "mvc_8:\tmvc\t0(8,%r2),0(%r3)\n"
>> -    "\tbr\t%r14");
>> +DEFINE_ASM_FUNCTION(mvc_8, "mvc 0(8,%r2),0(%r3)")
>> +
>> +extern const char return_from_main_1;
>>   
>>   static void safe_puts(const char *s)
>>   {
>> @@ -49,8 +56,9 @@ static struct {
>>   
>>   static void handle_signal(int sig, siginfo_t *info, void *ucontext)
>>   {
>> +    int err, i, n_frames;
>> +    void *frames[16];
>>       void *page;
>> -    int err;
>>   
>>       if (sig != expected.sig) {
>>           safe_puts("[  FAILED  ] wrong signal");
>> @@ -86,6 +94,17 @@ static void handle_signal(int sig, siginfo_t
>> *info, void *ucontext)
>>       default:
>>           break;
>>       }
>> +
>> +    n_frames = backtrace(frames, sizeof(frames) /
>> sizeof(frames[0]));
>> +    for (i = 0; i < n_frames; i++) {
>> +        if (frames[i] == &return_from_main_1) {
>> +            break;
>> +        }
>> +    }
>> +    if (i == n_frames) {
>> +        safe_puts("[  FAILED  ] backtrace() is broken");
>> +        _exit(1);
>> +    }
>>   }
>>   
>>   static void check_sigsegv(void *func, enum exception exception,
>> @@ -122,7 +141,7 @@ static void check_sigsegv(void *func, enum
>> exception exception,
>>       assert(err == 0);
>>   }
>>   
>> -int main(void)
>> +int main_1(void)
>>   {
>>       struct sigaction act;
>>       int err;
>> @@ -138,7 +157,7 @@ int main(void)
>>       safe_puts("[ RUN      ] Operation exception");
>>       expected.sig = SIGILL;
>>       expected.addr = illegal_op;
>> -    expected.psw_addr = (unsigned long)after_illegal_op;
>> +    expected.psw_addr = (unsigned long)&after_illegal_op;
>>       expected.exception = exception_operation;
>>       illegal_op();
>>       safe_puts("[       OK ]");
>> @@ -163,3 +182,25 @@ int main(void)
>>   
>>       _exit(0);
>>   }
>> +
>> +/*
>> + * Define main() in assembly in order to test that unwinding from
>> signal
>> + * handlers until main() works. This way we can define a specific
>> point that
>> + * the unwinder should reach. This is also better than defining
>> main() in C
>> + * and using inline assembly to call main_1(), since it's not easy
>> to get all
>> + * the clobbers right.
>> + */
>> +
>> +DEFINE_ASM_FUNCTION(main,
>> +    "stmg %r14,%r15,112(%r15)\n"
>> +    ".cfi_offset 14,-48\n"
>> +    ".cfi_offset 15,-40\n"
>> +    "lay %r15,-160(%r15)\n"
>> +    ".cfi_def_cfa_offset 320\n"
>> +    "brasl %r14,main_1\n"
>> +    ".globl return_from_main_1\n"
>> +    "return_from_main_1:\n"
>> +    "lmg %r14,%r15,272(%r15)\n"
>> +    ".cfi_restore 15\n"
>> +    ".cfi_restore 14\n"
>> +    ".cfi_def_cfa_offset 160");
> 
> Ping.

Acked-by: Thomas Huth <thuth@redhat.com>

Laurent, do you want to take these two patches through your linux-user 
branch, or shall I take them via the s390x branch?

  Thomas