mbox series

[(PING),0/1] target/riscv: misa to ISA string conversion fix

Message ID cover.1648270894.git.research_trasio@irq.a4lg.com (mailing list archive)
Headers show
Series target/riscv: misa to ISA string conversion fix | expand

Message

Tsukasa OI March 26, 2022, 5:01 a.m. UTC
[This is the same patch as previous ones]
<https://lists.nongnu.org/archive/html/qemu-riscv/2022-02/msg00098.html>
 (qemu-riscv only)
<https://lists.nongnu.org/archive/html/qemu-riscv/2022-02/msg00097.html>
 (resent due to configuration error of my mail server; qemu-riscv only)

I hope this is applied before the QEMU 7.0 release.

Signed-off-by: Tsukasa OI <research_trasio@irq.a4lg.com>
Reviewed-by: Frank Chang <frank.chang@sifive.com>

S and U are misa bits but not extensions (instead, they are supported
privilege modes).  Thus, they should not be copied to the ISA string.

I am truly surprised that this patchset is the THIRD attempt to fix this
longstanding problem.

(1) August 2019: by Palmer Dabbelt
<https://lists.nongnu.org/archive/html/qemu-riscv/2019-08/msg00165.html>
<https://lists.nongnu.org/archive/html/qemu-riscv/2019-08/msg00141.html>
<https://lists.nongnu.org/archive/html/qemu-riscv/2019-08/msg00259.html>

(2) April 2021: by Emmanuel Blot
<https://lists.nongnu.org/archive/html/qemu-riscv/2021-04/msg00248.html>

(3) February 2022: by me (this patchset)

I feel this is urgent to eliminate this bug now considering it required
a workaround to RISC-V Linux kernel as I pointed out:
<http://lists.infradead.org/pipermail/linux-riscv/2022-February/012252.html>


Though my patchset is first developed independently, this submitted
version is influenced by (2) Emmanuel Blot's patchset.  Thanks to this,
constant "[n]" can now be variable "[]".

It also fixes an ordering issue where 'C' should be preceded by 'L'
(order: 'L' -> 'C') as per the RISC-V ISA Manual (version 20191213),
Table 27.1.

It clarifies the role of `riscv_exts'.  It's a single-letter extrension
ordering list.




Tsukasa OI (1):
  target/riscv: misa to ISA string conversion fix

 target/riscv/cpu.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)


base-commit: f345abe36527a8b575482bb5a0616f43952bf1f4

Comments

Alistair Francis March 27, 2022, 11:29 p.m. UTC | #1
On Sat, Mar 26, 2022 at 3:46 PM Tsukasa OI <research_trasio@irq.a4lg.com> wrote:
>
> [This is the same patch as previous ones]

Hello,

Thanks for the patch!

> <https://lists.nongnu.org/archive/html/qemu-riscv/2022-02/msg00098.html>
>  (qemu-riscv only)

This never made it to the QEMU mailing list

> <https://lists.nongnu.org/archive/html/qemu-riscv/2022-02/msg00097.html>
>  (resent due to configuration error of my mail server; qemu-riscv only)

and neither did this

>
> I hope this is applied before the QEMU 7.0 release.

Unfortunately you have missed the window for 7.0. This patch will need
to be reviewed then applied for the next QEMU release.

>
> Signed-off-by: Tsukasa OI <research_trasio@irq.a4lg.com>
> Reviewed-by: Frank Chang <frank.chang@sifive.com>
>
> S and U are misa bits but not extensions (instead, they are supported
> privilege modes).  Thus, they should not be copied to the ISA string.
>
> I am truly surprised that this patchset is the THIRD attempt to fix this
> longstanding problem.

I'm sorry you feel this way, but this is the first time this patch has
been sent to the list since 2019.

I'm not sure why (1) wasn't applied, but (2) and (3) don't appear to
have been sent to the QEMU mailing list.

The separate RISC-V mailing list is confusing, but patches should be
sent to qemu-devel, as described at:
https://wiki.qemu.org/Contribute/MailingLists

Alistair

>
> (1) August 2019: by Palmer Dabbelt
> <https://lists.nongnu.org/archive/html/qemu-riscv/2019-08/msg00165.html>
> <https://lists.nongnu.org/archive/html/qemu-riscv/2019-08/msg00141.html>
> <https://lists.nongnu.org/archive/html/qemu-riscv/2019-08/msg00259.html>
>
> (2) April 2021: by Emmanuel Blot
> <https://lists.nongnu.org/archive/html/qemu-riscv/2021-04/msg00248.html>
>
> (3) February 2022: by me (this patchset)
>
> I feel this is urgent to eliminate this bug now considering it required
> a workaround to RISC-V Linux kernel as I pointed out:
> <http://lists.infradead.org/pipermail/linux-riscv/2022-February/012252.html>
>
>
> Though my patchset is first developed independently, this submitted
> version is influenced by (2) Emmanuel Blot's patchset.  Thanks to this,
> constant "[n]" can now be variable "[]".
>
> It also fixes an ordering issue where 'C' should be preceded by 'L'
> (order: 'L' -> 'C') as per the RISC-V ISA Manual (version 20191213),
> Table 27.1.
>
> It clarifies the role of `riscv_exts'.  It's a single-letter extrension
> ordering list.
>
>
>
>
> Tsukasa OI (1):
>   target/riscv: misa to ISA string conversion fix
>
>  target/riscv/cpu.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
>
>
> base-commit: f345abe36527a8b575482bb5a0616f43952bf1f4
> --
> 2.32.0
>
>
Tsukasa OI March 28, 2022, 1:09 p.m. UTC | #2
On 2022/03/28 8:29, Alistair Francis wrote:
> On Sat, Mar 26, 2022 at 3:46 PM Tsukasa OI <research_trasio@irq.a4lg.com> wrote:
>>
>> [This is the same patch as previous ones]
> 
> Hello,
> 
> Thanks for the patch!
> 
>> <https://lists.nongnu.org/archive/html/qemu-riscv/2022-02/msg00098.html>
>>  (qemu-riscv only)
> 
> This never made it to the QEMU mailing list
> 
>> <https://lists.nongnu.org/archive/html/qemu-riscv/2022-02/msg00097.html>
>>  (resent due to configuration error of my mail server; qemu-riscv only)
> 
> and neither did this

Ah, that was the reason!  I'm happy to know that.

> 
>>
>> I hope this is applied before the QEMU 7.0 release.
> 
> Unfortunately you have missed the window for 7.0. This patch will need
> to be reviewed then applied for the next QEMU release.

Understood.

> 
>>
>> Signed-off-by: Tsukasa OI <research_trasio@irq.a4lg.com>
>> Reviewed-by: Frank Chang <frank.chang@sifive.com>
>>
>> S and U are misa bits but not extensions (instead, they are supported
>> privilege modes).  Thus, they should not be copied to the ISA string.
>>
>> I am truly surprised that this patchset is the THIRD attempt to fix this
>> longstanding problem.
> 
> I'm sorry you feel this way, but this is the first time this patch has
> been sent to the list since 2019.
> 
> I'm not sure why (1) wasn't applied, but (2) and (3) don't appear to
> have been sent to the QEMU mailing list.
> 
> The separate RISC-V mailing list is confusing, but patches should be
> sent to qemu-devel, as described at:
> https://wiki.qemu.org/Contribute/MailingLists
> 
> Alistair

Thanks for letting me know.  From the next time, I will follow this.


> 
>>
>> (1) August 2019: by Palmer Dabbelt
>> <https://lists.nongnu.org/archive/html/qemu-riscv/2019-08/msg00165.html>
>> <https://lists.nongnu.org/archive/html/qemu-riscv/2019-08/msg00141.html>
>> <https://lists.nongnu.org/archive/html/qemu-riscv/2019-08/msg00259.html>
>>
>> (2) April 2021: by Emmanuel Blot
>> <https://lists.nongnu.org/archive/html/qemu-riscv/2021-04/msg00248.html>
>>
>> (3) February 2022: by me (this patchset)
>>
>> I feel this is urgent to eliminate this bug now considering it required
>> a workaround to RISC-V Linux kernel as I pointed out:
>> <http://lists.infradead.org/pipermail/linux-riscv/2022-February/012252.html>
>>
>>
>> Though my patchset is first developed independently, this submitted
>> version is influenced by (2) Emmanuel Blot's patchset.  Thanks to this,
>> constant "[n]" can now be variable "[]".
>>
>> It also fixes an ordering issue where 'C' should be preceded by 'L'
>> (order: 'L' -> 'C') as per the RISC-V ISA Manual (version 20191213),
>> Table 27.1.
>>
>> It clarifies the role of `riscv_exts'.  It's a single-letter extrension
>> ordering list.
>>
>>
>>
>>
>> Tsukasa OI (1):
>>   target/riscv: misa to ISA string conversion fix
>>
>>  target/riscv/cpu.c | 10 +++++-----
>>  1 file changed, 5 insertions(+), 5 deletions(-)
>>
>>
>> base-commit: f345abe36527a8b575482bb5a0616f43952bf1f4
>> --
>> 2.32.0
>>
>>
>