qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] qmp: Add 'memtranslate' QMP command


From: Markus Armbruster
Subject: Re: [PATCH] qmp: Add 'memtranslate' QMP command
Date: Wed, 31 Jul 2024 06:52:59 +0200
User-agent: Gnus/5.13 (Gnus v5.13)

junon@oro.sh writes:

> 31 July 2024 at 02:12, "Dr. David Alan Gilbert" <dave@treblig.org> wrote:
>
> Hello Dr. Gilbert,
>
>> 
>> * Josh Junon (junon@oro.sh) wrote:
>> 
>> Hi Josh,
>> 
>> > 
>> > This commit adds a new QMP/HMP command `memtranslate`,
>> > which translates a virtual address to a physical address
>> > using the guest's MMU.
>> >
>> > This uses the same mechanism that `[p]memsave` does to
>> > perform the translation.
>> >
>> > This commit also fixes a long standing issue of `[p]memsave`
>> > not properly handling higher-half virtual addresses correctly,
>> > namely when used over QMP/the monitor. The use and assumption of
>> > signed integers caused issues when parsing otherwise valid
>> > virtual addresses that instead caused signed integer overflow
>> > or ERANGE errors.
>> >
>> > Signed-off-by: Josh Junon <junon@oro.sh>
>> 
>> There's a few different changes in this one patch; so the first
>> thing is it needs splitting up; I suggest at least:
>> 
>>  a) Fixing the signedness problems
>> 
>>  b) The QMP implementation of the new command
>> 
>>  c) The HMP implementation of the new command
>> 
>> That would make it a lot easier to review - also, it's good
>> to get fixes in first!
>> Now, going back a step; how does this compare to the existing
>> 'gva2gpa' command which HMP has?
>> 
>
> Good catch, they're definitely the same. I didn't see that was there before, 
> perhaps because of the name. I've been looking for this exact command for a 
> while now, so it surprises me that I missed it!
>
> Since that's an HMP-only command, would it be okay if simply redirected its 
> definition to a new qmp_gva2gpa command so the implementation is all in one 
> spot?

If you have a use case for a QMP version, go right ahead.

> If that's amenable, I can patch in the signedness fixes, then submit 
> qmp_gva2gpa, then changing hmp_gva2gpa to use the qmp_gva2gpa similar to how 
> other HMP commands with QMP analogs are implemented. Just let me know if that 
> works and I'll get on it.

Sounds like a plan to me.

> I appreciate the response!
>
>
> Josh




reply via email to

[Prev in Thread] Current Thread [Next in Thread]