答复: 答复: [PATCH 1/1] Skip flatview_simplify() for cpu vendor zhaoxin

[FelixCui-oc]

hi paolo,

>So removing flatview_simplify() works because the higher area (0x10000

>and above) remains the same.  I guess the simplest thing to do is to
>apply flatview_simplify() only to I/O areas, though we can also consider
>removing it completely.  I'm not sure in which case it would provide a
>noticeable improvement.

Which one do you think is better? I prefer to remove flatview_simplify() directly.

Please help give some suggestions , and then i will resubmit the patch.

Best regards

FelixCui-oc

---

发件人: FelixCui-oc
发送时间: 2020年10月22日 14:31:26
收件人: Paolo Bonzini; Alex Williamson
抄送: Richard Henderson; Eduardo Habkost; qemu-devel@nongnu.org; RockCui-oc; Tony W Wang-oc; CobeChen-oc
主题: 答复: 答复: [PATCH 1/1] Skip flatview_simplify() for cpu vendor zhaoxin

 

hi paolo,

>So removing flatview_simplify() works because the higher area (0x10000

>and above) remains the same.  I guess the simplest thing to do is to
>apply flatview_simplify() only to I/O areas, though we can also consider
>removing it completely.  I'm not sure in which case it would provide a
>noticeable improvement.

I agree with you very much. Both of these cases can solve this bug.

Thanks.

Best regards

Felixcui-oc

---

发件人: Paolo Bonzini <pbonzini@redhat.com>
发送时间: 2020年10月22日 11:30:37
收件人: FelixCui-oc; Alex Williamson
抄送: Richard Henderson; Eduardo Habkost; qemu-devel@nongnu.org; RockCui-oc; Tony W Wang-oc; CobeChen-oc
主题: Re: 答复: [PATCH 1/1] Skip flatview_simplify() for cpu vendor zhaoxin

 

On 22/10/20 05:02, FelixCui-oc wrote:
> In addition, before write pam registers, flatview_simplify() has merged
> a very large range.For example,
>
> this large range is 0xc0000-0xbfffffff. So even if EHCI is configured to
> not allocate buffers in low memory,
>
> this bug will still occur.Thanks.

So removing flatview_simplify() works because the higher area (0x10000
and above) remains the same.  I guess the simplest thing to do is to
apply flatview_simplify() only to I/O areas, though we can also consider
removing it completely.  I'm not sure in which case it would provide a
noticeable improvement.

Paolo