libx86/CPUID: fix (not just) leaf 7 processing
PATCH 1/1 libx86/CPUID: fix (not just) leaf 7 processing
For one, subleaves within the respective union shouldn't live in separate sub-structures. And then x86_cpuid_policy_fill_native() should, as it did originally, iterate over all subleaves here as well as over all main leaves. Switch to using a "<= MIN()"-based approach similar to that used in x86_cpuid_copy_to_buffer(). Also follow this for the extended main leaves then.
xen/include/xen/lib/x86/cpuid.h +1 -2
@@ -181,8 +181,7 @@ struct cpuid_policyuint32_t _7d0; struct { DECL_BITFIELD(7d0); }; }; - }; - struct { +/* Subleaf 1. */ union {uint32_t _7a1;
xen/lib/x86/cpuid.c +6 -6
@@ -71,8 +71,8 @@ void x86_cpuid_policy_fill_native(structunsigned int i;cpuid_leaf (0 , &p->basic.raw[0 ]); - for ( i = 1; i < min_t(unsigned int, ARRAY_SIZE(p->basic.raw), - p->basic.max_leaf); ++i ) +for ( i =1 ; i <=MIN (p->basic.max_leaf, +ARRAY_SIZE (p->basic.raw) -1 ); ++i ) {switch ( i ) {
@@ -116,8 +116,8 @@ void x86_cpuid_policy_fill_native(struct {cpuid_count_leaf (7 ,0 , &p->feat.raw[0 ]); - for ( i = 1; i < min_t(unsigned int, ARRAY_SIZE(p->feat.raw), - p->feat.max_subleaf); ++i ) +for ( i =1 ; i <=MIN (p->feat.max_subleaf, +ARRAY_SIZE (p->feat.raw) -1 ); ++i )cpuid_count_leaf (7 , i, &p->feat.raw[i]); }
@@ -172,8 +172,8 @@ void x86_cpuid_policy_fill_native(struct/* Extended leaves. */ cpuid_leaf (0x80000000 , &p->extd.raw[0 ]); - for ( i = 1; i < min_t(unsigned int, ARRAY_SIZE(p->extd.raw), - p->extd.max_leaf +1 -0x80000000 ); ++i ) +for ( i =1 ; i <=MIN (p->extd.max_leaf &0xffffU , +ARRAY_SIZE (p->extd.raw) -1 ); ++i )cpuid_leaf (0x80000000 + i, &p->extd.raw[i]);x86_cpuid_policy_recalc_synth (p);
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Quoted 2 lines · Jan Beulich
For one, subleaves within the respective union shouldn't live in separate sub-structures.
Oops, and of course this stays hidden right now because there is no overlap in known bits between subleaf 0 and 1 yet.
Quoted 3 lines
And then
x86_cpuid_policy_fill_native()should,
as it did originally, iterate over all subleaves here as well as over all main leaves.
I was clearly asleep when writing that bugfix. Again - it says hidden because our local max_leaves are lower than ~all hardware these days.
Quoted 7 lines
Switch to using a "<= MIN()"-based approach similar to
that used inx86_cpuid_copy_to_buffer(). Also follow this for the extended main leaves then.Fixes 1bd2b750537b ("libx86: Fix 32bit stubdom build of x86_cpuid_policy_fill_native()")Fixes 97e4ebdcd765 ("x86/CPUID: support leaf 7 subleaf 1 / AVX512_BF16")Signed-off-by Jan Beulich <[email protected]>