x86/pv: Drop assertions from svm_load_segs()
PATCH 1/1 x86/pv: Drop assertions from svm_load_segs()
OSSTest has shown an assertion failure:
http://logs.test-lab.xenproject.org/osstest/logs/153906/test-xtf-amd64-amd64-1/serial-rimava1.log
These assertions were never appropriate, as they rule out legal (and, as it turns out, sensible perf-wise) inputs based on an expectation of how the sole caller would behave.
---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Wei Liu <[email protected]>
--- xen/arch/x86/hvm/svm/svm.c | 2 -- 1 file changed, 2 deletions(-)
xen/arch/x86/hvm/svm/svm.c +0 -2
@@ -1559,13 +1559,11 @@ bool svm_load_segs(unsigned int ldt_ents, unsigned long ldt_base, vmcb->ldtr.base = ldt_base; } -ASSERT (!(fs_sel & ~3 )); vmcb->fs.sel = fs_sel; vmcb->fs.attr =0 ; vmcb->fs.limit =0 ; vmcb->fs.base = fs_base; -ASSERT (!(gs_sel & ~3 )); vmcb->gs.sel = gs_sel; vmcb->gs.attr =0 ; vmcb->gs.limit =0 ;
Signature · 1 line
2.11.0
Quoted 6 lines · Andrew Cooper
OSSTest has shown an assertion failure:
http://logs.test-lab.xenproject.org/osstest/logs/153906/test-xtf-amd64-amd64-1/serial-rimava1.logThese assertions were never appropriate, as they rule out legal (and, as it turns out, sensible perf-wise) inputs based on an expectation of how the sole caller would behave.
I wouldn't put it this way - not coming here was a requirement in the old logic, when we expected non-null selectors to remain in sync with the GDT/LDT. But yes, they needs to be dropped now.
Quoted 1 line
Fixes ad0fd291c5 ("x86/pv: Rewrite segment context switching from scratch")
With your S-o-b added
Jan
Quoted 10 lines · Jan Beulich
Quoted 6 lines · Andrew Cooper
OSSTest has shown an assertion failure:
http://logs.test-lab.xenproject.org/osstest/logs/153906/test-xtf-amd64-amd64-1/serial-rimava1.logThese assertions were never appropriate, as they rule out legal (and, as it turns out, sensible perf-wise) inputs based on an expectation of how the sole caller would behave.
I wouldn't put it this way - not coming here was a requirement in the old logic, when we expected non-null selectors to remain in sync with the GDT/LDT. But yes, they needs to be dropped now.
Urgh. That is a can of worms, as this path doesn't preserve the GDT attributes/limits.
Quoted 4 lines
Quoted 2 lines
Fixes ad0fd291c5 ("x86/pv: Rewrite segment context switching from scratch")With your S-o-b added
Reviewed-by Jan Beulich <[email protected]>
Oops. Thanks, but no - I retract this patch. Will submit a different one shortly.
~Andrew