qemu-devel
[Top][All Lists]
Advanced

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

Re: [Qemu-devel] [PATCH] fw_cfg: fix endianness in fw_cfg_data_mem_read(


From: Laszlo Ersek
Subject: Re: [Qemu-devel] [PATCH] fw_cfg: fix endianness in fw_cfg_data_mem_read() / _write()
Date: Wed, 31 Dec 2014 18:04:03 +0100
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.3.0

On 12/31/14 17:23, Paolo Bonzini wrote:
> 
> 
> On 31/12/2014 16:17, Peter Maydell wrote:
>> On 31 December 2014 at 14:07, Laszlo Ersek <address@hidden> wrote:
>>>> Of course this also renders the issue mostly moot -- if none of us can
>>>> test the code on a BE host, then that use case simply doesn't exist in
>>>> practice.
>> If you can give me a test image and a command line I can test
>> it on one of the PPC64 boxes in the GCC compile farm.
> 
> You can follow the qtest steps in the commit messages of
> 6c87e3d5967a1d731b5f591a8f0ee6c319c14ca8:
> 
> $ arm-softmmu/qemu-system-arm -M virt -machine accel=qtest \
>              -qtest stdio -uuid 4600cb32-38ec-4b2f-8acb-81c6ea54f2d8
> writew 0x9020008 0x0200
> readl 0x9020000
> 
> The readl should return "OK 0x000000004600cb32".

The -uuid switch parses the UUID string with qemu_uuid_parse() into the
qemu_uuid array, and the first four bytes are stored with
"%02hhx%02hhx%02hhx%02hhx"; ie. byte-wise.

The fw_cfg_init1() function adds these bytes as a string (== saves a
reference to qemu_uuid), with fw_cfg_add_bytes(). Good.

I looked at the "readl" command implementation in
qtest_process_command(). I don't know what it intends to do.

*If* it prints the number that the guest CPU sees immediately when it
performs the wide read, then it should print 0x0000000032cb0046, on both
big and little endian hosts; assuming a little endian guest.

Namely, the fw_cfg (sub)string in question is [0x46, 0x00, 0xcb, 0x32].
The device is big endian, and the register accessor function should
return the 0x4600cb32 host value in qemu. The guest CPU should see the
same byte array [0x46, 0x00, 0xcb, 0x32], whose direct interpretation in
the little endian guest is 0x32cb0046.

... I think this is a good test case for what we *think* should happen :)

I think the following will happen if Peter executes this test:
- With the current code, QEMU will print different values when the host
endianness changes.
- With the patch under discussion, QEMU will print 0x0000000032cb0046 on
both host endiannesses.

Thanks
Laszlo



reply via email to

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