Re: [PATCH net] net: stmmac: Stop using a single dma_map() for multiple descriptors

[Dmitry Baryshkov]

On Tue, Sep 10, 2024 at 03:43:23PM GMT, Suraj Jaiswal wrote:
> 
> 
> -----Original Message-----
> From: Andrew Halaney <ahalaney@xxxxxxxxxx> 
> Sent: Wednesday, September 4, 2024 3:47 AM
> To: Suraj Jaiswal (QUIC) <quic_jsuraj@xxxxxxxxxxx>
> Cc: Vinod Koul <vkoul@xxxxxxxxxx>; bhupesh.sharma@xxxxxxxxxx; Andy Gross <agross@xxxxxxxxxx>; Bjorn Andersson <andersson@xxxxxxxxxx>; Konrad Dybcio <konrad.dybcio@xxxxxxxxxx>; David S. Miller <davem@xxxxxxxxxxxxx>; Eric Dumazet <edumazet@xxxxxxxxxx>; Jakub Kicinski <kuba@xxxxxxxxxx>; Rob Herring <robh+dt@xxxxxxxxxx>; Krzysztof Kozlowski <krzysztof.kozlowski+dt@xxxxxxxxxx>; Conor Dooley <conor+dt@xxxxxxxxxx>; Alexandre Torgue <alexandre.torgue@xxxxxxxxxxx>; Jose Abreu <joabreu@xxxxxxxxxxxx>; Maxime Coquelin <mcoquelin.stm32@xxxxxxxxx>; netdev@xxxxxxxxxxxxxxx; linux-arm-msm@xxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; linux-stm32@xxxxxxxxxxxxxxxxxxxxxxxxxxxx; Prasad Sodagudi <psodagud@xxxxxxxxxxx>; Rob Herring <robh@xxxxxxxxxx>; kernel <kernel@xxxxxxxxxxx>
> Subject: Re: [PATCH net] net: stmmac: Stop using a single dma_map() for multiple descriptors
> 
> WARNING: This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros.
> 
> On Mon, Sep 02, 2024 at 03:24:36PM GMT, Suraj Jaiswal wrote:
> > Currently same page address is shared
> > between multiple buffer addresses and causing smmu fault for other 
> > descriptor if address hold by one descriptor got cleaned.
> > Allocate separate buffer address for each descriptor for TSO path so 
> > that if one descriptor cleared it should not clean other descriptor 
> > address.
> 
> I think maybe you mean something like:
> 
>     Currently in the TSO case a page is mapped with dma_map_single(), and then
>     the resulting dma address is referenced (and offset) by multiple
>     descriptors until the whole region is programmed into the descriptors.
> 
>     This makes it possible for stmmac_tx_clean() to dma_unmap() the first of the
>     already processed descriptors, while the rest are still being processed
>     by the DMA engine. This leads to an iommu fault due to the DMA engine using
>     unmapped memory as seen below:
> 
>     <insert splat>
> 
>     You can reproduce this easily by <reproduction steps>.
> 
>     To fix this, let's map each descriptor's memory reference individually.
>     This way there's no risk of unmapping a region that's still being
>     referenced by the DMA engine in a later descriptor.
> 
> That's a bit nitpicky wording wise, but your first sentence is hard for me to follow (buffer addresses seems to mean descriptor?). I think showing a splat and mentioning how to reproduce is always a bonus as well.

Please fix your email client. It is impossible to understand where is
your answer and where comes the quoted text by Andrew. Use text emails,
text quotation (single '>') and no Outlook splat at the top of the email
("Original Message" with all the emails, etc).

-- 
With best wishes
Dmitry