Re: [PATCH 2/3] hvf: implement guest debugging on Apple Silicon hosts

[Mads Ynddal]

> On 4 Nov 2022, at 19.41, francesco.cagnin@gmail.com wrote:
> 
> From: Francesco Cagnin <fcagnin@quarkslab.com>
> 
> Support is added for single-stepping, software breakpoints, hardware
> breakpoints and watchpoints. The code has been structured like the KVM
> counterpart (and many parts are basically identical).
> 
> Guests can be debugged through the gdbstub.
> 
> Signed-off-by: Francesco Cagnin <fcagnin@quarkslab.com>
> ---
> accel/hvf/hvf-accel-ops.c | 124 ++++++++++++++++++++++++
> accel/hvf/hvf-all.c       |  24 +++++
> cpu.c                     |   3 +
> include/sysemu/hvf.h      |  29 ++++++
> include/sysemu/hvf_int.h  |   1 +
> target/arm/hvf/hvf.c      | 194 +++++++++++++++++++++++++++++++++++++-
> 6 files changed, 374 insertions(+), 1 deletion(-)


I've been working on the exact same features just last week, and had it working 
just hours before you posted, but you beat me to it. I can see we have solved it
almost exactly the same way, so I won't post my patchset.

I can see you are missing support for SSTEP_NOIRQ. I've handled it like this:

diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index 5ff5778d55..8b96d2f320 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -343,7 +343,7 @@ static int hvf_accel_init(MachineState *ms)

 static int hvf_gdbstub_sstep_flags(void)
 {
-    return SSTEP_ENABLE;
+    return SSTEP_ENABLE | SSTEP_NOIRQ;
 }

 static void hvf_accel_class_init(ObjectClass *oc, void *data)
diff --git a/target/arm/hvf/hvf.c b/target/arm/hvf/hvf.c
index dbc3605f6d..964a4ecf8a 100644
--- a/target/arm/hvf/hvf.c
+++ b/target/arm/hvf/hvf.c
@@ -1331,7 +1331,7 @@ int hvf_vcpu_exec(CPUState *cpu)
     hv_return_t r;
     bool advance_pc = false;

-    if (hvf_inject_interrupts(cpu)) {
+    if (!(cpu->singlestep_enabled & SSTEP_NOIRQ) && 
hvf_inject_interrupts(cpu)) {
         return EXCP_INTERRUPT;
     }

You'll have to suppress the interrupts while you're single-stepping the code. 
Otherwise, you'll only be stepping a few times, and suddenly get taken to the
interrupt-handler.

What issues do you have with multi-core systems?