Bug 216103
| Summary: | commit f5ff79fddf0efecca538046b5cc20fb3ded2ec4f breaks breaks user land DMA | ||
|---|---|---|---|
| Product: | Memory Management | Reporter: | Mark Hounschell (markh) |
| Component: | Other | Assignee: | Andrew Morton (akpm) |
| Status: | NEW --- | ||
| Severity: | normal | CC: | dmarkh, gregkh, markh, saxophonebritish |
| Priority: | P1 | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Kernel Version: | 5.18 | Subsystem: | |
| Regression: | No | Bisected commit-id: | |
|
Description
Mark Hounschell
2022-06-09 12:48:16 UTC
On Fri, Jun 17, 2022 at 02:44:10PM +0000, bugzilla-daemon@kernel.org wrote: > https://bugzilla.kernel.org/show_bug.cgi?id=216103 > > Mark Hounschell (markh@compro.net) changed: > > What |Removed |Added > ---------------------------------------------------------------------------- > CC| |gregkh@linuxfoundation.org Please email the relevant developers and mailing lists involved, bugzilla is not going to work for this, sorry. greg k-h I sent Christoph Hellwig <hch@lst.de> and Robin Murphy <robin.murphy@arm.com> emails containing just what you see in this bug. They more or less blew me off so I files this bug. I am not currently subscribed to the linux kernel list so did not post this there. On Mon, Jun 20, 2022 at 05:18:09PM +0000, bugzilla-daemon@kernel.org wrote: > https://bugzilla.kernel.org/show_bug.cgi?id=216103 > > --- Comment #2 from Mark Hounschell (markh@compro.net) --- > I sent Christoph Hellwig <hch@lst.de> and Robin Murphy <robin.murphy@arm.com> > emails containing just what you see in this bug. They more or less blew me > off > so I files this bug. I am not currently subscribed to the linux kernel list > so > did not post this there. There is never a need to subscribe to kernel lists to post to them. Please cc: them and the lists. OK. Thanks. I did not know this. I have done what you recommended. On Tue, Jun 21, 2022 at 01:43:28PM +0000, bugzilla-daemon@kernel.org wrote: > https://bugzilla.kernel.org/show_bug.cgi?id=216103 > > --- Comment #4 from Mark Hounschell (markh@compro.net) --- > OK. Thanks. I did not know this. I have done what you recommended. Wait, this is for an out-of-tree driver, of course there is nothing we can do about this for obvious reasons. It's not a regression at all, sorry. greg k-h Ok, I understand what you are saying. But it looks like one can no longer allocate memory for DMA using dma_alloc_coherent and then safely MMAP that same memory to userland. Safely, meaning can't be swapped out. That seems like a regression. On Tue, Jun 21, 2022 at 07:35:30PM +0000, bugzilla-daemon@kernel.org wrote: > https://bugzilla.kernel.org/show_bug.cgi?id=216103 > > --- Comment #6 from Mark Hounschell (markh@compro.net) --- > Ok, I understand what you are saying. But it looks like one can no longer > allocate memory for DMA using dma_alloc_coherent and then safely MMAP that > same > memory to userland. Safely, meaning can't be swapped out. That seems like a > regression. Again, for out-of-tree kernel modules, there is nothing we can do at all, and there is never any "regression" for them as we do not count them as any type of functionality that we can even consider, for obvious reasons. greg k-h |