----- Ursprüngliche Mail -----
>> A simple (but ugly!) approach would be redirecting mmap() requests on CUSE
>> devices to /dev/mem.
>> hmm?
>
> what requests are you talking about given that at the moment the CUSE
> client interface (cuse_lowlevel_ops) does not expose mmap?
The mmap() call itself. Of course you need to touch code.
Maybe just cuse_lowlevel.c, maybe kernel too.
Thanks,
//richard
On 09/02/21 17:29, Richard Weinberger wrote:
> The mmap() call itself. Of course you need to touch code.
> Maybe just cuse_lowlevel.c, maybe kernel too.
A patch had been submitted some years ago, more than once, asking for
inclusion in the kernel, but facts are that at that time nobody in the
FUSE group considered it. So much work and time lost, unfortunately.
----- Ursprüngliche Mail -----
> On 09/02/21 17:29, Richard Weinberger wrote:
>> The mmap() call itself. Of course you need to touch code.
>> Maybe just cuse_lowlevel.c, maybe kernel too.
>
> A patch had been submitted some years ago, more than once, asking for
> inclusion in the kernel, but facts are that at that time nobody in the
> FUSE group considered it. So much work and time lost, unfortunately.
Well, I think having a generic mmap() for CUSE is hard to achieve.
Thanks,
//richard
On 09/02/21 17:50, Richard Weinberger wrote:
> Well, I think having a generic mmap() for CUSE is hard to achieve.
Hard or not it did work for what I can tell you. I was not the original
author but I certainly contributed with testing that patch. Just to be
clear, by "not considered" I meant both the patch and the request were
completely ignored with no answers at all, silence in other words.
On Tue, Feb 9, 2021 at 6:46 PM Luca Risolia
<[email protected]> wrote:
>
> On 09/02/21 17:50, Richard Weinberger wrote:
> > Well, I think having a generic mmap() for CUSE is hard to achieve.
>
> Hard or not it did work for what I can tell you. I was not the original
> author but I certainly contributed with testing that patch. Just to be
> clear, by "not considered" I meant both the patch and the request were
> completely ignored with no answers at all, silence in other words.
I don't have any bad feelings about CUSE mmap, it was definitely not a
personal thing. But it also wasn't something that many people were
requesting, and the silence was probably at about the time that I was
taking a leave from paid maintainership. So a number of factors.
Patches are still welcome, if it's something that is used in practice
(e.g. by Android) that's a plus point.
Thanks,
Miklos