Re: [PATCH for-next v3] RDMA/core: Avoid hmm_dma_map_alloc() for virtual DMA devices

[Leon Romanovsky <leon@kernel.org>]

On Sun, May 25, 2025 at 05:27:23AM +0800, Greg Sword wrote:
> On Sat, May 24, 2025 at 10:43 PM Daisuke Matsuda <dskmtsd@gmail.com> wrote:
> >
> > Drivers such as rxe, which use virtual DMA, must not call into the DMA
> > mapping core since they lack physical DMA capabilities. Otherwise, a NULL
> > pointer dereference is observed as shown below. This patch ensures the RDMA
> > core handles virtual and physical DMA paths appropriately.

<...>

> > +EXPORT_SYMBOL(ib_dma_virt_map_alloc);
> >  #endif /* CONFIG_INFINIBAND_VIRT_DMA */
> >
> 
> Your ODP patches have caused significant issues, including system
> instability. The latest version of your patches has led to critical
> failures in our environment. Due to these ongoing problems, we have
> decided that our system will no longer incorporate your patches going
> forward.

Please be civil and appreciate the work done by other people. Daisuke
invested a lot of effort to implement ODP which is very non-trivial part
of RDMA HW.

If you can do it better, just do it, we will be happy to merge your
patches too.

At the end, we are talking about -next branch, which has potential to be
unstable, so next time provide detailed bug report to support your claims.

In addition, it was me who broke RXE.

Thanks