diff mbox series

linux-user: Add missing /proc/cpuinfo fields for sparc

Message ID Z37rL5jVYpOk6ufc@p100 (mailing list archive)
State New
Headers show
Series linux-user: Add missing /proc/cpuinfo fields for sparc | expand

Commit Message

Helge Deller Jan. 8, 2025, 9:16 p.m. UTC
Add some missing fields which may be parsed by userspace
applications.

Signed-off-by: Helge Deller <deller@gmx.de>

Comments

Richard Henderson Jan. 8, 2025, 10:59 p.m. UTC | #1
On 1/8/25 13:16, Helge Deller wrote:
> Add some missing fields which may be parsed by userspace
> applications.
> 
> Signed-off-by: Helge Deller <deller@gmx.de>
> 
> diff --git a/linux-user/sparc/target_proc.h b/linux-user/sparc/target_proc.h
> index 3bb3134a47..172d089fed 100644
> --- a/linux-user/sparc/target_proc.h
> +++ b/linux-user/sparc/target_proc.h
> @@ -8,7 +8,23 @@
>   
>   static int open_cpuinfo(CPUArchState *cpu_env, int fd)
>   {
> +    int i, num_cpus;
> +
> +    num_cpus = sysconf(_SC_NPROCESSORS_ONLN);
> +
> +    dprintf(fd, "cpu\t\t: TI UltraSparc II QEMU (BlackBird)\n");

Obviously won't match with non-default cpu.

r~

> +    dprintf(fd, "fpu\t\t: UltraSparc II integrated FPU\n");
> +    dprintf(fd, "promlib\t\t: Version 3 Revision 17\n");
> +    dprintf(fd, "prom\t\t: 3.17.0\n");

Do we really want to pretend with any of these?



r~
Helge Deller Jan. 8, 2025, 11:21 p.m. UTC | #2
On 1/8/25 23:59, Richard Henderson wrote:
> On 1/8/25 13:16, Helge Deller wrote:
>> Add some missing fields which may be parsed by userspace
>> applications.
>>
>> Signed-off-by: Helge Deller <deller@gmx.de>
>>
>> diff --git a/linux-user/sparc/target_proc.h b/linux-user/sparc/target_proc.h
>> index 3bb3134a47..172d089fed 100644
>> --- a/linux-user/sparc/target_proc.h
>> +++ b/linux-user/sparc/target_proc.h
>> @@ -8,7 +8,23 @@
>>   static int open_cpuinfo(CPUArchState *cpu_env, int fd)
>>   {
>> +    int i, num_cpus;
>> +
>> +    num_cpus = sysconf(_SC_NPROCESSORS_ONLN);
>> +
>> +    dprintf(fd, "cpu\t\t: TI UltraSparc II QEMU (BlackBird)\n");
>
> Obviously won't match with non-default cpu.

OIC!
Well, this would need much more coding which I won't do now
(and without any benefit as I haven't seen a program yet which
really cares about that specific CPU info).

>> +    dprintf(fd, "fpu\t\t: UltraSparc II integrated FPU\n");
>> +    dprintf(fd, "promlib\t\t: Version 3 Revision 17\n");
>> +    dprintf(fd, "prom\t\t: 3.17.0\n");
>
> Do we really want to pretend with any of these?

I'm now sparc expert, so I don't know.

My suggestion:
I change my patch to just add:

+    dprintf(fd, "ncpus probed\t: %d\n", num_cpus);
+    dprintf(fd, "ncpus active\t: %d\n", num_cpus);
+    dprintf(fd, "State:\n");
+    for (i = 0; i < num_cpus; i++) {
+        dprintf(fd, "CPU%d:\t\t: online\n", i);
+    }

Those were the lines which I was mostly interested in since
I found one application which wants to parse /proc/cpuinfo
to get number of CPUs in system.
Would that be acceptable (for now) ?

Helge
Richard Henderson Jan. 9, 2025, 12:40 a.m. UTC | #3
On 1/8/25 15:21, Helge Deller wrote:
> My suggestion:
> I change my patch to just add:
> 
> +    dprintf(fd, "ncpus probed\t: %d\n", num_cpus);
> +    dprintf(fd, "ncpus active\t: %d\n", num_cpus);
> +    dprintf(fd, "State:\n");
> +    for (i = 0; i < num_cpus; i++) {
> +        dprintf(fd, "CPU%d:\t\t: online\n", i);
> +    }
> 
> Those were the lines which I was mostly interested in since
> I found one application which wants to parse /proc/cpuinfo
> to get number of CPUs in system.
> Would that be acceptable (for now) ?

Yes, thanks.  If only linux had added a prctl or something to not require parsing...


r~
Helge Deller Jan. 9, 2025, 2:33 a.m. UTC | #4
On 1/9/25 01:40, Richard Henderson wrote:
> On 1/8/25 15:21, Helge Deller wrote:
>> My suggestion:
>> I change my patch to just add:
>>
>> +    dprintf(fd, "ncpus probed\t: %d\n", num_cpus);
>> +    dprintf(fd, "ncpus active\t: %d\n", num_cpus);
>> +    dprintf(fd, "State:\n");
>> +    for (i = 0; i < num_cpus; i++) {
>> +        dprintf(fd, "CPU%d:\t\t: online\n", i);
>> +    }
>>
>> Those were the lines which I was mostly interested in since
>> I found one application which wants to parse /proc/cpuinfo
>> to get number of CPUs in system.
>> Would that be acceptable (for now) ?
>
> Yes, thanks.  If only linux had added a prctl or something to not require parsing...

True, but most people forget about glibc providing sysconf(_SC_NPROCESSORS_ONLN)
as portable function so that people do not need to implement parsing
/proc/cpuinfo for each architecture themselves.

I'll respin the patch.

Helge
diff mbox series

Patch

diff --git a/linux-user/sparc/target_proc.h b/linux-user/sparc/target_proc.h
index 3bb3134a47..172d089fed 100644
--- a/linux-user/sparc/target_proc.h
+++ b/linux-user/sparc/target_proc.h
@@ -8,7 +8,23 @@ 
 
 static int open_cpuinfo(CPUArchState *cpu_env, int fd)
 {
+    int i, num_cpus;
+
+    num_cpus = sysconf(_SC_NPROCESSORS_ONLN);
+
+    dprintf(fd, "cpu\t\t: TI UltraSparc II QEMU (BlackBird)\n");
+    dprintf(fd, "fpu\t\t: UltraSparc II integrated FPU\n");
+    dprintf(fd, "promlib\t\t: Version 3 Revision 17\n");
+    dprintf(fd, "prom\t\t: 3.17.0\n");
     dprintf(fd, "type\t\t: sun4u\n");
+    dprintf(fd, "ncpus probed\t: %d\n", num_cpus);
+    dprintf(fd, "ncpus active\t: %d\n", num_cpus);
+    dprintf(fd, "MMU Type\t: Spitfire\n");
+    dprintf(fd, "State:\n");
+    for (i = 0; i < num_cpus; i++) {
+        dprintf(fd, "CPU%d:\t\t: online\n", i);
+    }
+
     return 0;
 }
 #define HAVE_ARCH_PROC_CPUINFO