On 2019/9/27 21:19, Peter Maydell wrote:
> On Fri, 6 Sep 2019 at 09:33, Xiang Zheng wrote:
>>
>> From: Dongjiu Geng
>>
>> kvm_hwpoison_page_add() and kvm_unpoison_all() will both be used by X86
>> and ARM platforms, so moving them into "include/sysemu/kvm_int.h" to
>> avoid duplicate code.
>>
>> Signed-off-by: Dongjiu Geng
>> Signed-off-by: Xiang Zheng
>> ---
>> accel/kvm/kvm-all.c | 33 +
>> include/sysemu/kvm_int.h | 23 +++
>> target/arm/kvm.c | 3 +++
>> target/i386/kvm.c| 34 --
>> 4 files changed, 59 insertions(+), 34 deletions(-)
>
>> static uint32_t adjust_ioeventfd_endianness(uint32_t val, uint32_t size)
>> {
>> #if defined(HOST_WORDS_BIGENDIAN) != defined(TARGET_WORDS_BIGENDIAN)
>> diff --git a/include/sysemu/kvm_int.h b/include/sysemu/kvm_int.h
>> index 72b2d1b3ae..3ad49f9a28 100644
>> --- a/include/sysemu/kvm_int.h
>> +++ b/include/sysemu/kvm_int.h
>> @@ -41,4 +41,27 @@ typedef struct KVMMemoryListener {
>> void kvm_memory_listener_register(KVMState *s, KVMMemoryListener *kml,
>>AddressSpace *as, int as_id);
>>
>> +/**
>> + * kvm_hwpoison_page_add:
>> + *
>> + * Parameters:
>> + * @ram_addr: the address in the RAM for the poisoned page
>> + *
>> + * Add a poisoned page to the list
>> + *
>> + * Return: None.
>> + */
>> +void kvm_hwpoison_page_add(ram_addr_t ram_addr);
>> +
>> +/**
>> + * kvm_unpoison_all:
>> + *
>> + * Parameters:
>> + * @param: some data may be passed to this function
>> + *
>> + * Free and remove all the poisoned pages in the list
>> + *
>> + * Return: None.
>> + */
>> +void kvm_unpoison_all(void *param);
>> #endif
>> diff --git a/target/arm/kvm.c b/target/arm/kvm.c
>> index b2eaa50b8d..3a110be7b8 100644
>> --- a/target/arm/kvm.c
>> +++ b/target/arm/kvm.c
>> @@ -20,6 +20,7 @@
>> #include "sysemu/sysemu.h"
>> #include "sysemu/kvm.h"
>> #include "sysemu/kvm_int.h"
>> +#include "sysemu/reset.h"
>> #include "kvm_arm.h"
>> #include "cpu.h"
>> #include "trace.h"
>> @@ -195,6 +196,8 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
>>
>> cap_has_mp_state = kvm_check_extension(s, KVM_CAP_MP_STATE);
>>
>> +qemu_register_reset(kvm_unpoison_all, NULL);
>> +
>
> Rather than registering the same reset handler in
> all the architectures, we could register it in the
> generic kvm_init() function. (For architectures that
> don't use the poison-list functionality the reset handler
> will harmlessly do nothing, because there will be nothing
> in the list.)
>
> This would allow you to not have to make the
> kvm_unpoison_all() function global -- it can be static
> in accel/tcg/kvm-all.c.
OK, I will move the register code into the kvm_init() function.
>
>> return 0;
>> }
>
> thanks
> -- PMM
>
> .
>
--
Thanks,
Xiang