Skip to contentSkip to conversation

tools/xen-cpuid: show enqcmd

PATCH In review Olaf Hering 3 participants 6 messages Jul 30, 2020 – Jul 31, 2020

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tools/xen-cpuid: show enqcmd

Translate <29> into a feature string.

---

not compile nor runtime tested.

 tools/misc/xen-cpuid.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/misc/xen-cpuid.c +1 -1
index ac3548dcfe..2446941a47 100644
@@ -133,7 +133,7 @@ static const char *const str_7c0[32] =
     [22] = "rdpid",
     /* 24 */                   [25] = "cldemote",
     /* 26 */                   [27] = "movdiri",
-    [28] = "movdir64b",
+    [28] = "movdir64b",        [29] = "enqcmd",
     [30] = "sgx-lc",
 };
 
Acked-by on v1 1/1
#1 lore raw
Quoted 3 lines · Olaf Hering

Translate <29> into a feature string.

Albeit I'm pretty sure there are more missing than just this lone one.

Jan

Quoted 12 lines
tools/misc/xen-cpuid.c +1 -1
@@ -133,7 +133,7 @@ static const char *const str_7c0[32] =
     [22] = "rdpid",
     /* 24 */                   [25] = "cldemote",
     /* 26 */                   [27] = "movdiri",
-    [28] = "movdir64b",
+    [28] = "movdir64b",        [29] = "enqcmd",
     [30] = "sgx-lc",
 };
 
#3 lore raw

Am Fri, 31 Jul 2020 13:04:35 +0100
schrieb Andrew Cooper <[email protected]>:

Quoted 2 lines

And in particular, probably missing from libxl_cpuid.c, which I was meaning to check when I've got a free moment.

Will a ever domU see this flag? I just spotted the <29> when comparing 'xen-cpuid' output between recent Xen releases. It shows up just in the 'Known' section at this point.

Olaf

#4 lore raw
Quoted 6 lines · Olaf Hering

Am Fri, 31 Jul 2020 13:04:35 +0100
schrieb Andrew Cooper <[email protected]>:

Quoted 2 lines

And in particular, probably missing from libxl_cpuid.c, which I was meaning to check when I've got a free moment.

Will a ever domU see this flag? I just spotted the <29> when comparing 'xen-cpuid' output between recent Xen releases. It shows up just in the 'Known' section at this point.

For future platforms supporting Compute eXpress Link, PCIPassthrough will be able to give various accelerators to VMs, and the ENQCMD{,S} instructions is what the guest will use to talk to hardware.

As a set of functionality, think DPDK but with with userspace (or the guest) able to submit work to hardware directly, through an interface designed to be safe for this kind of thing.

~Andrew

#5 lore raw
In reply to Andrew Cooper
Quoted 11 lines · Andrew Cooper
Quoted 7 lines · Jan Beulich
Quoted 3 lines · Olaf Hering

Translate <29> into a feature string.

Albeit I'm pretty sure there are more missing than just this lone one.

And in particular, probably missing from libxl_cpuid.c, which I was meaning to check when I've got a free moment.

As it's not just this one, but e.g. also the two movdir* ones, I thought I'd not require the other side to also be changed, the more that enqcmd also doesn't get exposed to guests at all right now.

Jan