qemu-devel
[Top][All Lists]
Advanced

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

Re: [PATCH v2 0/3] bulk: Remove legacy cpu_physical_memory_rw() API


From: Philippe Mathieu-Daudé
Subject: Re: [PATCH v2 0/3] bulk: Remove legacy cpu_physical_memory_rw() API
Date: Tue, 10 Dec 2024 16:10:57 +0100
User-agent: Mozilla Thunderbird

On 10/12/24 16:03, Peter Maydell wrote:
On Tue, 10 Dec 2024 at 15:00, Peter Maydell <peter.maydell@linaro.org> wrote:

On Tue, 10 Dec 2024 at 14:01, Philippe Mathieu-Daudé <philmd@linaro.org> wrote:

cpu_physical_memory_rw() API is documented as legacy
since 2017 (commit b7ecba0f6f6). Replace it by a more
recent API. Noticed while discussing with Peter:
e979b3ba-e701-4ac6-962a-19e4598ba947@linaro.org">https://lore.kernel.org/qemu-devel/e979b3ba-e701-4ac6-962a-19e4598ba947@linaro.org

I'm not sure we want to do this as a bulk automated
transformation, because in each case there is likely
a better thing we can do with the call than to use
address_space_memory. For example most of the uses in
devices probably want to have the device have an
AddressSpace property that the board wires up.

Also, examining each use gives us an opportunity to
consider the error handling (cpu_physical_memory_*()
drop errors silently) and whether there's an appropriate
MemTxAttrs we want to use.

Yes I noticed that and agree, but do we really want to improve
these devices? They have been using the legacy API for 7 years
without caring much.

I can repost split in 50 patches, hoping a dozen get merged
directly. But then I expect discussions requiring too much
unimportant work to happen, and the series being abandoned,
giving this legacy API 10 more years...

I'm being a bit negative so I'll post a v2, and then we can
open GitLab issues for devices needing to be improved.

Regards,

Phil.



reply via email to

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