Date:   Fri, 16 Sep 2022 01:51:09 +0800
From:   Vivian Wang <dramforever@...e.com>
To:     Conor.Dooley@...rochip.com, abrestic@...osinc.com
Cc:     palmer@...belt.com, paul.walmsley@...ive.com,
        coelacanthus@...look.com, c141028@...il.com,
        linux-riscv@...ts.infradead.org, linux-kernel@...r.kernel.org,
        atishp@...osinc.com, dramforever@...e.com
Subject: Re: [PATCH v3 2/2] riscv: Allow PROT_WRITE-only mmap()

On 9/16/22 01:40, Conor.Dooley@...rochip.com wrote:
> On 15/09/2022 18:27, Andrew Bresticker wrote:
>> On Thu, Sep 15, 2022 at 12:56 PM <Conor.Dooley@...rochip.com> wrote:
>>> On 09/09/2022 22:27, Andrew Bresticker wrote:
>>>> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
>>>>
>>>> Commit 2139619bcad7 ("riscv: mmap with PROT_WRITE but no PROT_READ is
>>>> invalid") made mmap() return EINVAL if PROT_WRITE was set wihtout
>>>> PROT_READ with the justification that a write-only PTE is considered a
>>>> reserved PTE permission bit pattern in the privileged spec. This check
>>>> is unnecessary since we let VM_WRITE imply VM_READ on RISC-V, and it is
>>>> inconsistent with other architectures that don't support write-only PTEs,
>>>> creating a potential software portability issue. Just remove the check
>>>> altogether and let PROT_WRITE imply PROT_READ as is the case on other
>>>> architectures.
>>>>
>>>> Note that this also allows PROT_WRITE|PROT_EXEC mappings which were
>>>> disallowed prior to the aforementioned commit; PROT_READ is implied in
>>>> such mappings as well.
>>>>
>>>> Fixes: 2139619bcad7 ("riscv: mmap with PROT_WRITE but no PROT_READ is invalid")
>>> For the naive members of the audience such as myself, this patch
>>> came after a non-fixes patch in the series. What is the dependence
>>> of this patch on the other one (if any)?
>> This patch is dependent on the first. Happy to re-spin with a "Fixes"
>> tag on the first patch (or maybe Palmer can add when applying).
> If it is a fix, then it should have a fixes tag. If it's cosmetic reorg
> to make the fix easier then no & it should be moved after the fix. If
> it is neither then you should prob mention it in the cover or under the
> --- /shrug

Basically what happens fixes-wise is that patch 1 fixes the original
problem in a different way, and patch 2 undoes the previous patch (with
small additional fixes). IMO this needs a cover to explain what's going
on to those who missed the v1 and v2 discussion thread.

v1 here: https://lore.kernel.org/all/20220908170133.1159747-1-abrestic@rivosinc.com

Thanks,
dram

> Thanks,
> Conor.
>
>>>> Reviewed-by: Atish Patra <atishp@...osinc.com>
>>>> Signed-off-by: Andrew Bresticker <abrestic@...osinc.com>
>>>> ---
>>>> v1 -> v2: Update access_error() to account for write-implies-read
>>>> v2 -> v3: Separate into two commits
>>>> ---
>>>>  arch/riscv/kernel/sys_riscv.c | 3 ---
>>>>  1 file changed, 3 deletions(-)
>>>>
>>>> diff --git a/arch/riscv/kernel/sys_riscv.c b/arch/riscv/kernel/sys_riscv.c
>>>> index 571556bb9261..5d3f2fbeb33c 100644
>>>> --- a/arch/riscv/kernel/sys_riscv.c
>>>> +++ b/arch/riscv/kernel/sys_riscv.c
>>>> @@ -18,9 +18,6 @@ static long riscv_sys_mmap(unsigned long addr, unsigned long len,
>>>>         if (unlikely(offset & (~PAGE_MASK >> page_shift_offset)))
>>>>                 return -EINVAL;
>>>>
>>>> -       if (unlikely((prot & PROT_WRITE) && !(prot & PROT_READ)))
>>>> -               return -EINVAL;
>>>> -
>>>>         return ksys_mmap_pgoff(addr, len, prot, flags, fd,
>>>>                                offset >> (PAGE_SHIFT - page_shift_offset));
>>>>  }
>>>> --
>>>> 2.25.1
>>>>
>>>>
>>>> _______________________________________________
>>>> linux-riscv mailing list
>>>> linux-riscv@...ts.infradead.org
>>>> http://lists.infradead.org/mailman/listinfo/linux-riscv

Powered by blists - more mailing lists