[PATCH v3 0/2] linux-user/s390x: signal with SIGFPE on compare-and-trap

[PATCH v3 0/2] linux-user/s390x: signal with SIGFPE on compare-and-trap

[Jonathan Albrecht]

qemu-s390x signals with SIGILL on compare-and-trap instructions. This
breaks OpenJDK which expects SIGFPE in its implementation of implicit
exceptions.

This patch depends on [PATCH v6 0/2] target/s390x: Fix SIGILL and SIGFPE
psw.addr reporting
https://lore.kernel.org/qemu-devel/20210705210434.45824-1-iii@linux.ibm.com/

Based-on: 20210705210434.45824-1-iii@linux.ibm.com

v1 -> v2:
- Update to latest version of '... psw.addr reporting' patch
- Rebase to master and fix conflicts in tests/tcg/s390x/Makefile.target

v2 -> v3:
- Check for non-simulated IEEE exception DXC codes explicitly when
  getting si_codes
- Ensure si_code is set in all cases
- Improve comments

Jonathan Albrecht (2):
  linux-user/s390x: signal with SIGFPE on compare-and-trap
  tests/tcg: Test that compare-and-trap raises SIGFPE

 linux-user/s390x/cpu_loop.c     |  54 ++++++++++-------
 tests/tcg/s390x/Makefile.target |   1 +
 tests/tcg/s390x/trap.c          | 102 ++++++++++++++++++++++++++++++++
 3 files changed, 137 insertions(+), 20 deletions(-)
 create mode 100644 tests/tcg/s390x/trap.c

-- 
2.31.1

[PATCH v3 1/2] linux-user/s390x: signal with SIGFPE on compare-and-trap

[Jonathan Albrecht]

Currently when a compare-and-trap instruction is executed, qemu will
always raise a SIGILL signal. On real hardware, a SIGFPE is raised.

Change the PGM_DATA case in cpu_loop to follow the behavior in
linux kernel /arch/s390/kernel/traps.c.
 * Only raise SIGILL if DXC == 0
 * If DXC matches a non-simulated IEEE exception, raise SIGFPE with
   correct si_code
 * Raise SIGFPE with si_code == 0 for everything else

When applied on 20210705210434.45824-2-iii@linux.ibm.com, this fixes
crashes in the java jdk such as the linked bug.

Buglink: https://bugs.launchpad.net/qemu/+bug/1920913
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/319
Signed-off-by: Jonathan Albrecht <jonathan.albrecht@linux.vnet.ibm.com>
---
 linux-user/s390x/cpu_loop.c | 54 +++++++++++++++++++++++--------------
 1 file changed, 34 insertions(+), 20 deletions(-)

diff --git a/linux-user/s390x/cpu_loop.c b/linux-user/s390x/cpu_loop.c
index 22f2e89c62..6a69a6dd26 100644
--- a/linux-user/s390x/cpu_loop.c
+++ b/linux-user/s390x/cpu_loop.c
@@ -25,6 +25,35 @@
 /* s390x masks the fault address it reports in si_addr for SIGSEGV and SIGBUS */
 #define S390X_FAIL_ADDR_MASK -4096LL
 
+static int get_pgm_data_si_code(int dxc_code)
+{
+    switch (dxc_code) {
+    /* Non-simulated IEEE exceptions */
+    case 0x80:
+        return TARGET_FPE_FLTINV;
+    case 0x40:
+        return TARGET_FPE_FLTDIV;
+    case 0x20:
+    case 0x28:
+    case 0x2c:
+        return TARGET_FPE_FLTOVF;
+    case 0x10:
+    case 0x18:
+    case 0x1c:
+        return TARGET_FPE_FLTUND;
+    case 0x08:
+    case 0x0c:
+        return TARGET_FPE_FLTRES;
+    }
+    /*
+     * Non-IEEE and simulated IEEE:
+     * Includes compare-and-trap, quantum exception, etc.
+     * Simulated IEEE are included here to match current
+     * s390x linux kernel.
+     */
+    return 0;
+}
+
 void cpu_loop(CPUS390XState *env)
 {
     CPUState *cs = env_cpu(env);
@@ -106,29 +135,14 @@ void cpu_loop(CPUS390XState *env)
 
             case PGM_DATA:
                 n = (env->fpc >> 8) & 0xff;
-                if (n == 0xff) {
-                    /* compare-and-trap */
+                if (n == 0) {
                     goto do_sigill_opn;
-                } else {
-                    /* An IEEE exception, simulated or otherwise.  */
-                    if (n & 0x80) {
-                        n = TARGET_FPE_FLTINV;
-                    } else if (n & 0x40) {
-                        n = TARGET_FPE_FLTDIV;
-                    } else if (n & 0x20) {
-                        n = TARGET_FPE_FLTOVF;
-                    } else if (n & 0x10) {
-                        n = TARGET_FPE_FLTUND;
-                    } else if (n & 0x08) {
-                        n = TARGET_FPE_FLTRES;
-                    } else {
-                        /* ??? Quantum exception; BFP, DFP error.  */
-                        goto do_sigill_opn;
-                    }
-                    sig = TARGET_SIGFPE;
-                    goto do_signal_pc;
                 }
 
+                sig = TARGET_SIGFPE;
+                n = get_pgm_data_si_code(n);
+                goto do_signal_pc;
+
             default:
                 fprintf(stderr, "Unhandled program exception: %#x\n", n);
                 cpu_dump_state(cs, stderr, 0);
-- 
2.31.1

[PATCH v3 2/2] tests/tcg: Test that compare-and-trap raises SIGFPE

[Jonathan Albrecht]

Signed-off-by: Jonathan Albrecht <jonathan.albrecht@linux.vnet.ibm.com>
---
 tests/tcg/s390x/Makefile.target |   1 +
 tests/tcg/s390x/trap.c          | 102 ++++++++++++++++++++++++++++++++
 2 files changed, 103 insertions(+)
 create mode 100644 tests/tcg/s390x/trap.c

diff --git a/tests/tcg/s390x/Makefile.target b/tests/tcg/s390x/Makefile.target
index 0a5b25c156..d440ecd6f7 100644
--- a/tests/tcg/s390x/Makefile.target
+++ b/tests/tcg/s390x/Makefile.target
@@ -9,6 +9,7 @@ TESTS+=exrl-trtr
 TESTS+=pack
 TESTS+=mvo
 TESTS+=mvc
+TESTS+=trap
 
 # This triggers failures on s390x hosts about 4% of the time
 run-signals: signals
diff --git a/tests/tcg/s390x/trap.c b/tests/tcg/s390x/trap.c
new file mode 100644
index 0000000000..d4c61c7f52
--- /dev/null
+++ b/tests/tcg/s390x/trap.c
@@ -0,0 +1,102 @@
+/*
+ * Copyright 2021 IBM Corp.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or (at
+ * your option) any later version. See the COPYING file in the top-level
+ * directory.
+ */
+
+#include <stdarg.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <errno.h>
+#include <string.h>
+#include <signal.h>
+
+static void error1(const char *filename, int line, const char *fmt, ...)
+{
+    va_list ap;
+    va_start(ap, fmt);
+    fprintf(stderr, "%s:%d: ", filename, line);
+    vfprintf(stderr, fmt, ap);
+    fprintf(stderr, "\n");
+    va_end(ap);
+    exit(1);
+}
+
+static int __chk_error(const char *filename, int line, int ret)
+{
+    if (ret < 0) {
+        error1(filename, line, "%m (ret=%d, errno=%d/%s)",
+               ret, errno, strerror(errno));
+    }
+    return ret;
+}
+
+#define error(fmt, ...) error1(__FILE__, __LINE__, fmt, ## __VA_ARGS__)
+
+#define chk_error(ret) __chk_error(__FILE__, __LINE__, (ret))
+
+int sigfpe_count;
+int sigill_count;
+
+static void sig_handler(int sig, siginfo_t *si, void *puc)
+{
+    if (sig == SIGFPE) {
+        if (si->si_code != 0) {
+            error("unexpected si_code: 0x%x != 0", si->si_code);
+        }
+        ++sigfpe_count;
+        return;
+    }
+
+    if (sig == SIGILL) {
+        ++sigill_count;
+        return;
+    }
+
+    error("unexpected signal 0x%x\n", sig);
+}
+
+int main(int argc, char **argv)
+{
+    sigfpe_count = sigill_count = 0;
+
+    struct sigaction act;
+
+    /* Set up SIG handler */
+    act.sa_sigaction = sig_handler;
+    sigemptyset(&act.sa_mask);
+    act.sa_flags = SA_SIGINFO;
+    chk_error(sigaction(SIGFPE, &act, NULL));
+    chk_error(sigaction(SIGILL, &act, NULL));
+
+    uint64_t z = 0x0ull;
+    uint64_t lz = 0xffffffffffffffffull;
+    asm volatile (
+        "lg %%r13,%[lz]\n"
+        "cgitne %%r13,0\n" /* SIGFPE */
+        "lg %%r13,%[z]\n"
+        "cgitne %%r13,0\n" /* no trap */
+        "nopr\n"
+        "lg %%r13,%[lz]\n"
+        "citne %%r13,0\n" /* SIGFPE */
+        "lg %%r13,%[z]\n"
+        "citne %%r13,0\n" /* no trap */
+        "nopr\n"
+        :
+        : [z] "m" (z), [lz] "m" (lz)
+        : "memory", "r13");
+
+    if (sigfpe_count != 2) {
+        error("unexpected SIGFPE count: %d != 2", sigfpe_count);
+    }
+    if (sigill_count != 0) {
+        error("unexpected SIGILL count: %d != 0", sigill_count);
+    }
+
+    printf("PASS\n");
+    return 0;
+}
-- 
2.31.1

Re: [PATCH v3 1/2] linux-user/s390x: signal with SIGFPE on compare-and-trap

[Richard Henderson]

On 7/9/21 9:04 AM, Jonathan Albrecht wrote:
> Currently when a compare-and-trap instruction is executed, qemu will
> always raise a SIGILL signal. On real hardware, a SIGFPE is raised.
> 
> Change the PGM_DATA case in cpu_loop to follow the behavior in
> linux kernel /arch/s390/kernel/traps.c.
>   * Only raise SIGILL if DXC == 0
>   * If DXC matches a non-simulated IEEE exception, raise SIGFPE with
>     correct si_code
>   * Raise SIGFPE with si_code == 0 for everything else
> 
> When applied on20210705210434.45824-2-iii@linux.ibm.com, this fixes
> crashes in the java jdk such as the linked bug.
> 
> Buglink:https://bugs.launchpad.net/qemu/+bug/1920913
> Resolves:https://gitlab.com/qemu-project/qemu/-/issues/319
> Signed-off-by: Jonathan Albrecht<jonathan.albrecht@linux.vnet.ibm.com>
> ---
>   linux-user/s390x/cpu_loop.c | 54 +++++++++++++++++++++++--------------
>   1 file changed, 34 insertions(+), 20 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

Re: [PATCH v3 0/2] linux-user/s390x: signal with SIGFPE on compare-and-trap

[Laurent Vivier]

Le 09/07/2021 à 18:04, Jonathan Albrecht a écrit :
> qemu-s390x signals with SIGILL on compare-and-trap instructions. This
> breaks OpenJDK which expects SIGFPE in its implementation of implicit
> exceptions.
> 
> This patch depends on [PATCH v6 0/2] target/s390x: Fix SIGILL and SIGFPE
> psw.addr reporting
> https://lore.kernel.org/qemu-devel/20210705210434.45824-1-iii@linux.ibm.com/
> 
> Based-on: 20210705210434.45824-1-iii@linux.ibm.com
> 
> v1 -> v2:
> - Update to latest version of '... psw.addr reporting' patch
> - Rebase to master and fix conflicts in tests/tcg/s390x/Makefile.target
> 
> v2 -> v3:
> - Check for non-simulated IEEE exception DXC codes explicitly when
>   getting si_codes
> - Ensure si_code is set in all cases
> - Improve comments
> 
> Jonathan Albrecht (2):
>   linux-user/s390x: signal with SIGFPE on compare-and-trap
>   tests/tcg: Test that compare-and-trap raises SIGFPE
> 
>  linux-user/s390x/cpu_loop.c     |  54 ++++++++++-------
>  tests/tcg/s390x/Makefile.target |   1 +
>  tests/tcg/s390x/trap.c          | 102 ++++++++++++++++++++++++++++++++
>  3 files changed, 137 insertions(+), 20 deletions(-)
>  create mode 100644 tests/tcg/s390x/trap.c
> 

Series applied to my linux-user-for-6.1 branch.

Thanks,
Laurent

Re: [PATCH v3 0/2] linux-user/s390x: signal with SIGFPE on compare-and-trap

[jonathan.albrecht]

On 2021-07-12 4:02 pm, Laurent Vivier wrote:
> Le 09/07/2021 à 18:04, Jonathan Albrecht a écrit :
>> qemu-s390x signals with SIGILL on compare-and-trap instructions. This
>> breaks OpenJDK which expects SIGFPE in its implementation of implicit
>> exceptions.
>> 
>> This patch depends on [PATCH v6 0/2] target/s390x: Fix SIGILL and 
>> SIGFPE
>> psw.addr reporting
>> https://lore.kernel.org/qemu-devel/20210705210434.45824-1-iii@linux.ibm.com/
>> 
>> Based-on: 20210705210434.45824-1-iii@linux.ibm.com
>> 
>> 
> 
> Series applied to my linux-user-for-6.1 branch.
> 

Thanks Laurent, I see this series has been applied to 
https://github.com/vivier/qemu/commits/linux-user-for-6.1 but the 
following series that this is based on also needs to be applied:

https://lore.kernel.org/qemu-devel/20210705210434.45824-1-iii@linux.ibm.com/

Did some local testing and looks like missing that series caused 
https://app.travis-ci.com/github/vivier/qemu/jobs/523853464 to fail.

Oh, just saw Ilya's email that the test patch has not been reviewed. 
Hopefully that can happen so they can both make it in.

Thanks,

Jon

Re: [PATCH v3 0/2] linux-user/s390x: signal with SIGFPE on compare-and-trap

[Laurent Vivier]

Le 12/07/2021 à 23:29, jonathan.albrecht a écrit :
> On 2021-07-12 4:02 pm, Laurent Vivier wrote:
>> Le 09/07/2021 à 18:04, Jonathan Albrecht a écrit :
>>> qemu-s390x signals with SIGILL on compare-and-trap instructions. This
>>> breaks OpenJDK which expects SIGFPE in its implementation of implicit
>>> exceptions.
>>>
>>> This patch depends on [PATCH v6 0/2] target/s390x: Fix SIGILL and SIGFPE
>>> psw.addr reporting
>>> https://lore.kernel.org/qemu-devel/20210705210434.45824-1-iii@linux.ibm.com/
>>>
>>> Based-on: 20210705210434.45824-1-iii@linux.ibm.com
>>>
>>>
>>
>> Series applied to my linux-user-for-6.1 branch.
>>
> 
> Thanks Laurent, I see this series has been applied to
> https://github.com/vivier/qemu/commits/linux-user-for-6.1 but the following series that this is
> based on also needs to be applied:
> 
> https://lore.kernel.org/qemu-devel/20210705210434.45824-1-iii@linux.ibm.com/
> 
> Did some local testing and looks like missing that series caused
> https://app.travis-ci.com/github/vivier/qemu/jobs/523853464 to fail.
> 
> Oh, just saw Ilya's email that the test patch has not been reviewed. Hopefully that can happen so
> they can both make it in.

I've removed these two patches from my patch queue. I'll do a new one with both series when they
will be ready.

Thanks,
Laurent