x86_seg_* uses architectural encodings. Therefore, we can fold the prefix
handling cases together and derive the segment from the prefix byte itself.
No functional change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
---
CC: Jan Beulich <JBeulich@suse.com>
CC: Roger Pau Monné <roger.pau@citrix.com>
This makes a marginal saving of 47 bytes, indicating (not unexpectedly) that
the optimiser cannot fold the case statements automatically.
fs/gs is weirder. The expression is 4 + (b & 1), and for some reason this
adds +264 bytes to the function. Even if the logical expression is larger
than two simple stores (and it probably is), it's not +264 bytes larger...
---
xen/arch/x86/x86_emulate/decode.c | 19 +++++--------------
1 file changed, 5 insertions(+), 14 deletions(-)
diff --git a/xen/arch/x86/x86_emulate/decode.c b/xen/arch/x86/x86_emulate/decode.c
index de836068fdd8..ee4cbdc0002c 100644
--- a/xen/arch/x86/x86_emulate/decode.c
+++ b/xen/arch/x86/x86_emulate/decode.c
@@ -1043,17 +1043,12 @@ int x86emul_decode(struct x86_emulate_state *s,
case 0x67: /* address-size override */
ad_bytes = def_ad_bytes ^ (mode_64bit() ? 12 : 6);
break;
- case 0x2e: /* CS override / ignored in 64-bit mode */
+ case 0x26: /* ES override */
+ case 0x2e: /* CS override */
+ case 0x36: /* SS override */
+ case 0x3e: /* DS override, all ignored in 64-bit mode */
if ( !mode_64bit() )
- override_seg = x86_seg_cs;
- break;
- case 0x3e: /* DS override / ignored in 64-bit mode */
- if ( !mode_64bit() )
- override_seg = x86_seg_ds;
- break;
- case 0x26: /* ES override / ignored in 64-bit mode */
- if ( !mode_64bit() )
- override_seg = x86_seg_es;
+ override_seg = (b >> 3) & 3;
break;
case 0x64: /* FS override */
override_seg = x86_seg_fs;
@@ -1061,10 +1056,6 @@ int x86emul_decode(struct x86_emulate_state *s,
case 0x65: /* GS override */
override_seg = x86_seg_gs;
break;
- case 0x36: /* SS override / ignored in 64-bit mode */
- if ( !mode_64bit() )
- override_seg = x86_seg_ss;
- break;
case 0xf0: /* LOCK */
s->lock_prefix = true;
break;
--
2.30.2
On 11.04.2024 17:23, Andrew Cooper wrote: > x86_seg_* uses architectural encodings. Therefore, we can fold the prefix > handling cases together and derive the segment from the prefix byte itself. > > No functional change. > > Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> I notice we already have suitable BUILD_BUG_ON()s, as we use similar logic already for PUSH/POP of the selector registers. Jan
© 2016 - 2024 Red Hat, Inc.