On 30/08/2023 2:30 pm, Jan Beulich wrote:
> On 24.08.2023 17:26, Jinoh Kang wrote:
>> From: Andrew Cooper <andrew.coop...@citrix.com>
>>
>> All #DB exceptions result in an update of %dr6, but this isn't captured in
>> Xen's handling.
>>
>> PV guests generally work by modifying %dr6 before raising #DB, whereas HVM
>> guests do nothing and have a single-step special case in the lowest levels of
>> {vmx,svm}_inject_event(). All of this is buggy, but in particular, task
>> switches with the trace flag never end up signalling BT in %dr6.
>>
>> To begin resolving this issue, add a new pending_dbg field to x86_event
>> (unioned with cr2 to avoid taking any extra space), and introduce
>> {pv,hvm}_inject_debug_exn() helpers to replace the current callers using
>> {pv,hvm}_inject_hw_exception().
>>
>> A key property is that pending_dbg is taken with positive polarity to deal
>> with RTM sensibly. Most callers pass in a constant, but callers passing in a
>> hardware %dr6 value need to xor the value with X86_DR6_DEFAULT to flip the
>> polarity of RTM and reserved fields.
>>
>> For PV guests, move the ad-hoc updating of %dr6 into pv_inject_event(). This
>> in principle breaks the handing of RTM in do_debug(), but PV guests can't
>> actually enable MSR_DEBUGCTL.RTM yet, so this doesn't matter in practice.
>>
>> Signed-off-by: Andrew Cooper <andrew.coop...@citrix.com>
>> Reviewed-by: Roger Pau Monné <roger....@citrix.com>
>> Reviewed-by: Jan Beulich <jbeul...@suse.com>
>> [ jinoh: Rebase onto staging, forward declare struct vcpu ]
>> Signed-off-by: Jinoh Kang <jinoh.kang...@gmail.com>
>> ---
>> CC: Andrew Cooper <andrew.coop...@citrix.com>
>> CC: Jan Beulich <jbeul...@suse.com>
>> CC: Wei Liu <w...@xen.org>
>> CC: Roger Pau Monné <roger....@citrix.com>
>> CC: Jun Nakajima <jun.nakaj...@intel.com>
>> CC: Kevin Tian <kevin.t...@intel.com>
>> CC: Tim Deegan <t...@xen.org>
>> CC: Tamas K Lengyel <ta...@tklengyel.com>
>> CC: Alexandru Isaila <aisa...@bitdefender.com>
>> CC: Petre Pircalabu <ppircal...@bitdefender.com>
>>
>> v1 -> v2: [S-o-b fixes. More details below.]
>>
>> - Update DR6 for gdbsx when trapped in PV guest kernel mode
> I take it that this refers to ...
>
>> --- a/xen/arch/x86/traps.c
>> +++ b/xen/arch/x86/traps.c
>> @@ -1989,17 +1989,17 @@ void do_debug(struct cpu_user_regs *regs)
>> return;
>> }
>>
>> - /* Save debug status register where guest OS can peek at it */
>> - v->arch.dr6 |= (dr6 & ~X86_DR6_DEFAULT);
>> - v->arch.dr6 &= (dr6 | ~X86_DR6_DEFAULT);
>> -
>> if ( guest_kernel_mode(v, regs) && v->domain->debugger_attached )
>> {
>> + /* Save debug status register where gdbsx can peek at it */
>> + v->arch.dr6 |= (dr6 & ~X86_DR6_DEFAULT);
>> + v->arch.dr6 &= (dr6 | ~X86_DR6_DEFAULT);
>> +
>> domain_pause_for_debugger();
>> return;
>> }
> ... this code movement. I'm afraid this should have resulted in you
> dropping the earlier R-b, and I'm further afraid I'm not convinced
> this is correct, despite seeing why you would want to do this. The
> issue is that this way you also alter guest-visible state, when the
> intention is that such now happen via ...
>
>> - pv_inject_hw_exception(X86_EXC_DB, X86_EVENT_NO_EC);
>> + pv_inject_debug_exn(dr6 ^ X86_DR6_DEFAULT);
>> }
> ... this call alone. I fear I can't currently see how to get both
> aspects right, other than by breaking up
I think it was wrongly broken up in my RFC series too.
With hindsight, I think the series wants rearranging to introduce
x86_merge_dr6() first, and then fix up PV and HVM separately. They're
independent logic paths.
~Andrew
