Bug#1091858: [PATCH] mips/math-emu: fix emulation of the prefx instruction
W dniu 7.01.2025 o 08:28, Salvatore Bonaccorso pisze:
> Hi Mateusz,
>
> On Fri, Jan 03, 2025 at 10:14:47PM +0100, Mateusz Jo??czyk wrote:
>> The above patch fixes the issue for me (on 32-bit mipsel), but would need some further testing.
>>
>> I wrote:
>>
>>> Fixes: 51061b8876a3 ("MIPS: math-emu: Fix prefx detection and COP1X function field definition")
>> but I realized this has been broken by
>> commit 919af8b96c89 ("MIPS: Make definitions of MIPSInst_FMA_{FUNC,FMTM} consistent with MIPS64 manual")
>> in v5.12 which modified the MIPSInst_FMA_FFMT macro without updating the users.
> I'm possibly bit confused. 919af8b96c89 ("MIPS: Make definitions of
> MIPSInst_FMA_{FUNC,FMTM} consistent with MIPS64 manual") in 5.12-rc1
> is not backported to the 6.1.y series.
>
> What am I missing here?
>
> Regards,
> Salvatore
Hello,
I wasn't clear enough.
In my original patch, I added:
> Fixes: 51061b8876a3 ("MIPS: math-emu: Fix prefx detection and COP1X function field definition")
I was mistaken: The commit 51061b8876a3 was added before v3.14-rc8 ( https://kernel.dance/#51061b8876a3 ), and it was not at fault.
The problem was caused by
commit 919af8b96c89 ("MIPS: Make definitions of MIPSInst_FMA_{FUNC,FMTM} consistent with MIPS64 manual")
which was introduced in v5.12-rc1 ( https://kernel.dance/#919af8b96c89 ) .
Between v3.14-rc8 and v5.11 everything was apparently working OK. It was in v5.12-rc1 when the problems started.
Linux v5.12-rc1 introduced the bug and it was present in all later kernels, including v6.1.y.
There was no need for backporting, just natural version progression.
I hope that this explains everything and I understood you correctly.
I don't know why the bug was not reported earlier, perhaps some earlier versions of Debian
did not use zstd for initramfs compression.
Greetings,
Mateusz
Reply to: