Commit 48ad4222 authored by Mingwei Zhang's avatar Mingwei Zhang Committed by Sean Christopherson

KVM: selftests: Enable checking on xcomp_bv in amx_test

After tilerelease instruction, AMX tiles are in INIT state. According to
Intel SDM vol 1. 13.10: "If RFBM[i] = 1, XSTATE_BV[i] is set to the
value of XINUSE[i].", XSTATE_BV[18] should be cleared after xsavec.

On the other hand, according to Intel SDM vol 1. 13.4.3: "If XCOMP_BV[i] =
1, state component i is located at a byte offset locationI from the base
address of the XSAVE area". Since at the time of xsavec, XCR0[18] is set
indicating AMX tile data component is still enabled, xcomp_bv[18] should be
set.

Complete the checks by adding the assert to xcomp_bv[18] after xsavec.
Signed-off-by: default avatarMingwei Zhang <mizhang@google.com>
Link: https://lore.kernel.org/r/20230221163655.920289-5-mizhang@google.comSigned-off-by: default avatarSean Christopherson <seanjc@google.com>
parent bec357a4
...@@ -197,6 +197,7 @@ static void __attribute__((__flatten__)) guest_code(struct tile_config *amx_cfg, ...@@ -197,6 +197,7 @@ static void __attribute__((__flatten__)) guest_code(struct tile_config *amx_cfg,
xstate->header.xstate_bv = XFEATURE_MASK_XTILEDATA; xstate->header.xstate_bv = XFEATURE_MASK_XTILEDATA;
__xsavec(xstate, XFEATURE_MASK_XTILEDATA); __xsavec(xstate, XFEATURE_MASK_XTILEDATA);
GUEST_ASSERT(!(xstate->header.xstate_bv & XFEATURE_MASK_XTILEDATA)); GUEST_ASSERT(!(xstate->header.xstate_bv & XFEATURE_MASK_XTILEDATA));
GUEST_ASSERT(xstate->header.xcomp_bv & XFEATURE_MASK_XTILEDATA);
/* xfd=0x40000, disable amx tiledata */ /* xfd=0x40000, disable amx tiledata */
wrmsr(MSR_IA32_XFD, XFEATURE_MASK_XTILEDATA); wrmsr(MSR_IA32_XFD, XFEATURE_MASK_XTILEDATA);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment