Query the SYS_HEAPINFO semicall and do some basic verification of the
information via libc calls.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

---
v5
  - drop asnprintf, just mix printf output
  - don't use brk() but do sanity check
---
 .../multiarch/arm-compat-semi/semihosting.c   | 44 ++++++++++++++++++-
 1 file changed, 43 insertions(+), 1 deletion(-)

diff --git a/tests/tcg/multiarch/arm-compat-semi/semihosting.c b/tests/tcg/multiarch/arm-compat-semi/semihosting.c
index b3fd16cd12..8e9cfb578c 100644
--- a/tests/tcg/multiarch/arm-compat-semi/semihosting.c
+++ b/tests/tcg/multiarch/arm-compat-semi/semihosting.c
@@ -8,9 +8,13 @@
  */
 
 #define SYS_WRITE0      0x04
+#define SYS_HEAPINFO    0x16
 #define SYS_REPORTEXC   0x18
 
 #include <stdint.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
 #include "semicall.h"
 
 int main(int argc, char *argv[argc])
@@ -21,8 +25,46 @@ int main(int argc, char *argv[argc])
     uintptr_t exit_block[2] = {0x20026, 0};
     uintptr_t exit_code = (uintptr_t) &exit_block;
 #endif
+    struct {
+        void *heap_base;
+        void *heap_limit;
+        void *stack_base;
+        void *stack_limit;
+    } info;
+    void *ptr_to_info = (void *) &info;
 
-    __semi_call(SYS_WRITE0, (uintptr_t) "Hello World");
+    __semi_call(SYS_WRITE0, (uintptr_t) "Checking HeapInfo\n");
+
+    memset(&info, 0, sizeof(info));
+    __semi_call(SYS_HEAPINFO, (uintptr_t) &ptr_to_info);
+
+    if (info.heap_base == NULL || info.heap_limit == NULL) {
+        printf("null heap: %p -> %p\n", info.heap_base, info.heap_limit);
+        exit(1);
+    } else if (info.heap_base != NULL && info.heap_limit != NULL) {
+        /* Error if heap base is above limit */
+        if ((uintptr_t) info.heap_base >= (uintptr_t) info.heap_limit) {
+            printf("heap base %p >= heap_limit %p\n",
+                   info.heap_base, info.heap_limit);
+            exit(2);
+        }
+    }
+
+    if (info.stack_base == NULL) {
+        printf("null stack: %p -> %p\n", info.stack_base, info.stack_limit);
+        exit(3);
+    } else if (info.stack_base != NULL && info.stack_limit != NULL) {
+        /* Error if stack base is below limit */
+        if ((uintptr_t) info.stack_base < (uintptr_t) info.stack_limit) {
+            printf("stack base %p < stack_limit %p\n",
+                   info.stack_base, info.stack_limit);
+            exit(4);
+        }
+    }
+    printf("heap: %p -> %p\n", info.heap_base, info.heap_limit);
+    printf("stack: %p -> %p\n", info.stack_base, info.stack_limit);
+
+    __semi_call(SYS_WRITE0, (uintptr_t) "Passed HeapInfo checks");
     __semi_call(SYS_REPORTEXC, exit_code);
     /* if we get here we failed */
     return -1;
-- 
2.20.1

On Fri, 12 Mar 2021 at 10:31, Alex Bennée <alex.bennee@linaro.org> wrote:
>
> Query the SYS_HEAPINFO semicall and do some basic verification of the
> information via libc calls.

Are we both testing system emulation and linux-user, or just one of the two?
(Not that I want to hold this up if we're only testing one, but coverage
of both, including M-profile, seems like it would be useful; cf also
https://bugs.launchpad.net/qemu/+bug/1918302 which is M-profile system-emulation
SYS_HEAPINFO being broken in 5.2.)

thanks
-- PMM

Peter Maydell <peter.maydell@linaro.org> writes:

> On Fri, 12 Mar 2021 at 10:31, Alex Benn=C3=A9e <alex.bennee@linaro.org> wro=
> te:
>>
>> Query the SYS_HEAPINFO semicall and do some basic verification of the
>> information via libc calls.
>
> Are we both testing system emulation and linux-user, or just one of the two=
> ?

Currently just linux-user. There is no inherent reason we couldn't add a
system emulation test but I felt that was already reasonably exercised
by your semihosting tests.

> (Not that I want to hold this up if we're only testing one, but coverage
> of both, including M-profile, seems like it would be useful; cf also
> https://bugs.launchpad.net/qemu/+bug/1918302 which is M-profile system-emul=
> ation
> SYS_HEAPINFO being broken in 5.2.)
>
> thanks
> -- PMM


-- 
Alex Bennée

On Fri, 12 Mar 2021 at 11:24, Alex Bennée <alex.bennee@linaro.org> wrote:
>
>
> Peter Maydell <peter.maydell@linaro.org> writes:
>
> > On Fri, 12 Mar 2021 at 10:31, Alex Benn=C3=A9e <alex.bennee@linaro.org> wro=
> > te:
> >>
> >> Query the SYS_HEAPINFO semicall and do some basic verification of the
> >> information via libc calls.
> >
> > Are we both testing system emulation and linux-user, or just one of the two=
> > ?
>
> Currently just linux-user. There is no inherent reason we couldn't add a
> system emulation test but I felt that was already reasonably exercised
> by your semihosting tests.

Do we run those automatically anywhere ? I agree that probably the
best thing is to get those into the CI loop somehow (and then put
a SYS_HEAPINFO test there.)

-- PMM

Peter Maydell <peter.maydell@linaro.org> writes:

> On Fri, 12 Mar 2021 at 11:24, Alex Benn=C3=A9e <alex.bennee@linaro.org> wro=
> te:
>>
>>
>> Peter Maydell <peter.maydell@linaro.org> writes:
>>
>> > On Fri, 12 Mar 2021 at 10:31, Alex Benn=3DC3=3DA9e <alex.bennee@linaro.=
> org> wro=3D
>> > te:
>> >>
>> >> Query the SYS_HEAPINFO semicall and do some basic verification of the
>> >> information via libc calls.
>> >
>> > Are we both testing system emulation and linux-user, or just one of the=
>  two=3D
>> > ?
>>
>> Currently just linux-user. There is no inherent reason we couldn't add a
>> system emulation test but I felt that was already reasonably exercised
>> by your semihosting tests.
>
> Do we run those automatically anywhere ? I agree that probably the
> best thing is to get those into the CI loop somehow (and then put
> a SYS_HEAPINFO test there.)

Not currently. I have toyed with series that make integrate the building
of external test suites into the source tree. There are a number that it
would be useful to have a short "make" away including:

  kvm-unit-tests
  LTP
  diy7 (memory model testing)
  sbsa-ref tools

and I guess the semihosting testd would be another good example.

-- 
Alex Bennée