Re: [Qemu-devel] [PATCH qemu v2] hmp: Print if memory section is registered with an accelerator

["Dr. David Alan Gilbert" <dgilbert@redhat.com>]

* Alexey Kardashevskiy (aik@ozlabs.ru) wrote:
> 
> 
> On 04/01/2019 04:37, Dr. David Alan Gilbert wrote:
> > * Alexey Kardashevskiy (aik@ozlabs.ru) wrote:
> >>
> >>
> >> On 17/12/2018 23:47, Philippe Mathieu-Daudé wrote:
> >>> On 12/17/18 2:27 AM, Alexey Kardashevskiy wrote:
> >>>> On 14/12/2018 22:07, Philippe Mathieu-Daudé wrote:
> >>>>> Hi Alexey,
> >>>>>
> >>>>> On 12/14/18 3:58 AM, Alexey Kardashevskiy wrote:
> >>>>>> This adds an accelerator name to the "into mtree -f" to tell the user if
> >>>>>> a particular memory section is registered with the accelerator;
> >>>>>> the primary user for this is KVM and such information is useful
> >>>>>> for debugging purposes.
> >>>>>>
> >>>>>> This adds a has_memory() callback to the accelerator class allowing any
> >>>>>> accelerator to have a label in that memory tree dump.
> >>>>>>
> >>>>>> Since memory sections are passed to memory listeners and get registered
> >>>>>> in accelerators (rather than memory regions), this only prints new labels
> >>>>>> for flatviews attached to the system address space.
> >>>>>>
> >>>>>> An example:
> >>>>>>  Root memory region: system
> >>>>>>   0000000000000000-0000002fffffffff (prio 0, ram): /objects/mem0 kvm
> >>>>>>   0000003000000000-0000005fffffffff (prio 0, ram): /objects/mem1 kvm
> >>>>>>   0000200000000020-000020000000003f (prio 1, i/o): virtio-pci
> >>>>>>   0000200080000000-000020008000003f (prio 0, i/o): capabilities
> >>>>>>
> >>>>>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> >>>>>> ---
> >>>>>>
> >>>>>> This supercedes "[PATCH qemu] hmp: Print if memory section is registered in KVM"
> >>>>>>
> >>>>>> ---
> >>>>>> Changes:
> >>>>>> v2:
> >>>>>> * added an accelerator callback instead of hardcoding it to kvm only
> >>>>>> ---
> >>>>>>  include/sysemu/accel.h |  2 ++
> >>>>>>  accel/kvm/kvm-all.c    | 10 ++++++++++
> >>>>>>  memory.c               | 22 ++++++++++++++++++++++
> >>>>>>  3 files changed, 34 insertions(+)
> >>>>>>
> >>>>>> diff --git a/include/sysemu/accel.h b/include/sysemu/accel.h
> >>>>>> index 637358f..30b456d 100644
> >>>>>> --- a/include/sysemu/accel.h
> >>>>>> +++ b/include/sysemu/accel.h
> >>>>>> @@ -25,6 +25,7 @@
> >>>>>>  
> >>>>>>  #include "qom/object.h"
> >>>>>>  #include "hw/qdev-properties.h"
> >>>>>> +#include "exec/hwaddr.h"
> >>>>>>  
> >>>>>>  typedef struct AccelState {
> >>>>>>      /*< private >*/
> >>>>>> @@ -41,6 +42,7 @@ typedef struct AccelClass {
> >>>>>>      int (*available)(void);
> >>>>>>      int (*init_machine)(MachineState *ms);
> >>>>>>      void (*setup_post)(MachineState *ms, AccelState *accel);
> >>>>>> +    bool (*has_memory)(MachineState *ms, hwaddr start_addr, hwaddr size);
> >>>>>>      bool *allowed;
> >>>>>>      /*
> >>>>>>       * Array of global properties that would be applied when specific
> >>>>>> diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
> >>>>>> index 4880a05..634f386 100644
> >>>>>> --- a/accel/kvm/kvm-all.c
> >>>>>> +++ b/accel/kvm/kvm-all.c
> >>>>>> @@ -2589,11 +2589,21 @@ int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target)
> >>>>>>      return r;
> >>>>>>  }
> >>>>>>  
> >>>>>> +static bool kvm_accel_has_memory(MachineState *ms, hwaddr start_addr,
> >>>>>> +                                 hwaddr size)
> >>>>>> +{
> >>>>>> +    KVMState *kvm = KVM_STATE(ms->accelerator);
> >>>>>> +    KVMMemoryListener *kml = &kvm->memory_listener;
> >>>>>> +
> >>>>>> +    return NULL != kvm_lookup_matching_slot(kml, start_addr, size);
> >>>>>> +}
> >>>>>> +
> >>>>>>  static void kvm_accel_class_init(ObjectClass *oc, void *data)
> >>>>>>  {
> >>>>>>      AccelClass *ac = ACCEL_CLASS(oc);
> >>>>>>      ac->name = "KVM";
> >>>>>>      ac->init_machine = kvm_init;
> >>>>>> +    ac->has_memory = kvm_accel_has_memory;
> >>>>>>      ac->allowed = &kvm_allowed;
> >>>>>>  }
> >>>>>>  
> >>>>>> diff --git a/memory.c b/memory.c
> >>>>>> index d14c6de..61e758a 100644
> >>>>>> --- a/memory.c
> >>>>>> +++ b/memory.c
> >>>>>> @@ -29,7 +29,9 @@
> >>>>>>  #include "exec/ram_addr.h"
> >>>>>>  #include "sysemu/kvm.h"
> >>>>>>  #include "sysemu/sysemu.h"
> >>>>>> +#include "sysemu/accel.h"
> >>>>>>  #include "hw/qdev-properties.h"
> >>>>>> +#include "hw/boards.h"
> >>>>>>  #include "migration/vmstate.h"
> >>>>>>  
> >>>>>>  //#define DEBUG_UNASSIGNED
> >>>>>> @@ -2924,6 +2926,8 @@ struct FlatViewInfo {
> >>>>>>      int counter;
> >>>>>>      bool dispatch_tree;
> >>>>>>      bool owner;
> >>>>>> +    AccelClass *ac;
> >>>>>> +    const char *ac_name;
> >>>>>>  };
> >>>>>>  
> >>>>>>  static void mtree_print_flatview(gpointer key, gpointer value,
> >>>>>> @@ -2939,6 +2943,7 @@ static void mtree_print_flatview(gpointer key, gpointer value,
> >>>>>>      int n = view->nr;
> >>>>>>      int i;
> >>>>>>      AddressSpace *as;
> >>>>>> +    bool system_as = false;
> >>>>>>  
> >>>>>>      p(f, "FlatView #%d\n", fvi->counter);
> >>>>>>      ++fvi->counter;
> >>>>>> @@ -2950,6 +2955,9 @@ static void mtree_print_flatview(gpointer key, gpointer value,
> >>>>>>              p(f, ", alias %s", memory_region_name(as->root->alias));
> >>>>>>          }
> >>>>>>          p(f, "\n");
> >>>>>> +        if (as == &address_space_memory) {
> >>>>>> +            system_as = true;
> >>>>>> +        }
> >>>>>>      }
> >>>>>>  
> >>>>>>      p(f, " Root memory region: %s\n",
> >>>>>> @@ -2985,6 +2993,13 @@ static void mtree_print_flatview(gpointer key, gpointer value,
> >>>>>>          if (fvi->owner) {
> >>>>>>              mtree_print_mr_owner(p, f, mr);
> >>>>>>          }
> >>>>>> +
> >>>>>> +        if (system_as && fvi->ac &&
> >>>>>> +            fvi->ac->has_memory(current_machine,
> >>>>>> +                                int128_get64(range->addr.start),
> >>>>>> +                                MR_SIZE(range->addr.size) + 1)) {
> >>>>>> +            p(f, " %s", fvi->ac_name);
> >>>>>
> >>>>> Why not simply display fvi->ac->name?
> >>>>> You could avoid to add the ac_name field.
> >>>>
> >>>>
> >>>> Well, I thought I better print whatever the user passed via the command
> >>>> line (which is current_machine->accel and equals to "kvm" in my case)
> >>>> rather than robotic, dry and excessive "kvm-accel".
> >>>
> >>> I have no hit for 'kvm-accel':
> >>>
> >>> $ git grep kvm-accel
> >>> $
> >>>
> >>> Names looks human friendly:
> >>
> >>
> >> Ah, I confused with object_class_get_name(). Anyway, I'd still show the
> >> user provided accelerator name than some internal name.
> > 
> > Aren't they exactly the same?
> 
> Nope. "kvm" vs. "kvm-accel".
> 
> > I think the accel= thing from the user is
> > parsed by comparing it with the ACCEL_CLASS_NAME declarations:
> > 
> > ./target/i386/whpx-all.c:    .name = ACCEL_CLASS_NAME("whpx"),
> > ./target/i386/hax-all.c:    .name = ACCEL_CLASS_NAME("hax"),
> > ./hw/xen/xen-common.c:#define TYPE_XEN_ACCEL ACCEL_CLASS_NAME("xen")
> > ./qtest.c:#define TYPE_QTEST_ACCEL ACCEL_CLASS_NAME("qtest")
> > ./accel/tcg/tcg-all.c:#define TYPE_TCG_ACCEL ACCEL_CLASS_NAME("tcg")
> > ./include/sysemu/hvf.h:#define TYPE_HVF_ACCEL ACCEL_CLASS_NAME("hvf")
> > ./include/sysemu/kvm_int.h:#define TYPE_KVM_ACCEL ACCEL_CLASS_NAME("kvm")
> > 
> > and the code in accel/accel.c accel_find
> 
> 
> https://git.qemu.org/?p=qemu.git;a=blob;f=include/sysemu/accel.h;h=5565e00a96136bc5024c17c401116ef6a497138a;hb=HEAD#l58
> 
>   55 #define TYPE_ACCEL "accel"
>   56
>   57 #define ACCEL_CLASS_SUFFIX  "-" TYPE_ACCEL
>   58 #define ACCEL_CLASS_NAME(a) (a ACCEL_CLASS_SUFFIX)

(Following up from your pings).
I'm OK from the HMP side, but I agree it would be nice to avoid adding
the ac_name field in FlatViewInfo if it's only reason is to get rid of
the '-accel'.   If that's the only reason why don't you just truncate
the fvi->ac->name when you display it?  You could do a build-time
check that TYPE_ACCEL and ACCEL_CLASS_SUFFIX haven't changed.

Dave

> 
> 
> > 
> > Dave
> > 
> >>
> >>>
> >>> $ git grep 'ac->name ='
> >>> accel/kvm/kvm-all.c:2595:    ac->name = "KVM";
> >>> accel/tcg/tcg-all.c:74:    ac->name = "tcg";
> >>> hw/xen/xen-common.c:184:    ac->name = "Xen";
> >>> target/i386/hax-all.c:1088:    ac->name = "HAX";
> >>> target/i386/hvf/hvf.c:959:    ac->name = "HVF";
> >>> target/i386/whpx-all.c:1477:    ac->name = "WHPX";
> >>> qtest.c:755:    ac->name = "QTest";
> >>>
> >>>>>
> >>>>>> +        }
> >>>>>>          p(f, "\n");
> >>>>>>          range++;
> >>>>>>      }
> >>>>>> @@ -3028,6 +3043,13 @@ void mtree_info(fprintf_function mon_printf, void *f, bool flatview,
> >>>>>>          };
> >>>>>>          GArray *fv_address_spaces;
> >>>>>>          GHashTable *views = g_hash_table_new(g_direct_hash, g_direct_equal);
> >>>>>> +        AccelClass *ac = ACCEL_GET_CLASS(current_machine->accelerator);
> >>>>>> +
> >>>>>> +        if (ac->has_memory) {
> >>>>>> +            fvi.ac = ac;
> >>>>>> +            fvi.ac_name = current_machine->accel ? current_machine->accel :
> >>>>>> +                object_class_get_name(OBJECT_CLASS(ac));
> >>>>>> +        }
> >>>>>>  
> >>>>>>          /* Gather all FVs in one table */
> >>>>>>          QTAILQ_FOREACH(as, &address_spaces, address_spaces_link) {
> >>>>>>
> >>
> >> -- 
> >> Alexey
> > --
> > Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK
> > 
> 
> -- 
> Alexey
--
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK