Hello folks,
Almost a year ago I sent a patchset to the Linux MIPS community. The main target of the patchset
was to get rid from the old bootmem allocator usage at the MIPS architecture. Additionally I had
a problem with CMA usage on my MIPS machine due to some struct page-related issue. Moving to the
memblock allocator fixed the problem and gave us benefits like smaller memory consumption,
powerful memblock API to be used within the arch code.
Sorry for the delay. I promised to get back to this work within two months since my last post, but
there are more over five months passed. Now when some free time finally available at my schedule
I am ready to go on with this patchset and finally finish it up. After a short glance at the old
code I suppose there should be some refactoring made. Particularly:
- completely get rid from the boot_mem_map[] array usage. After the memblock code utilization
is added, the boot_mem_map array will be excessive.
- unpin memblock/bootmem unrelated changes to a separate series so the patchset wouldn't be
that big.
- take into account all the comments the community users posted on my initial patchset.
Still there are several questions, I need to have answered before the development is started.
Question to the community in general. Are you still interested in this work to be done and the
patchset to be submitted for review?
@[email protected]. Last time you moved the patchset to the RFC status for some reason. I asked
you twice to send to me your comments about the patches, but obviously you didn't have time for
this, then you might have just forgotten. Could you please confirm whether you are interested in
these alterations and will be ready to review them?
@[email protected]. Could you confirm if you are still interested in the patchset and
ready to update the Loongson64 platform code so one would be compatible with it?
@[email protected]. Will you still be able to update the SGI IP27 code so one would work on top
of the patches?
@[email protected]. Do you still possess the Octeon MIPS64 platform to test the patchset?
Thanks folks for you willingness to help
Regards,
-Sergey
Hello Serge,
On 19/12/17 21:14, Serge Semin wrote:
> Almost a year ago I sent a patchset to the Linux MIPS community. The main target of the patchset
> was to get rid from the old bootmem allocator usage at the MIPS architecture. Additionally I had
> a problem with CMA usage on my MIPS machine due to some struct page-related issue. Moving to the
> memblock allocator fixed the problem and gave us benefits like smaller memory consumption,
> powerful memblock API to be used within the arch code.
[...]
> @[email protected]. Do you still possess the Octeon MIPS64 platform to test the patchset?
yes, I'd like to test such a serious change on Octeon2. Please include me in the distribution list.
--
Best regards,
Alexander Sverdlin.
Hello folks,
> > -----Original Message-----
> > From: Peter Wotton
> > Sent: 20 December 2017 16:18
> > To: 'Serge Semin'
> > Cc: Paul Burton; Matt Redfearn; '[email protected]'; Marcin
> > Nowakowski
> > Subject: RE: [RFC] MIPS memblock: Remove bootmem code and switch to
> > NO_BOOTMEM
> >
> > Hi Serge,
> > The imgtec.com email addresses have changed to mips.com following the
> > sale of the MIPS division, so forwarding email for Marcin, MattR, James and
> > Paul (altho' they'll have also received via the LMO list) Regards Peter.
> > Regards
> > Peter.
> >
Thanks Peter. I'll be sending the email to the @mips.com domain then.
> From: Alexander Sverdlin
> To: 'Serge Semin'
> Hello Serge,
>
> On 19/12/17 21:14, Serge Semin wrote:
> > Almost a year ago I sent a patchset to the Linux MIPS community. The main target of the patchset
> > was to get rid from the old bootmem allocator usage at the MIPS architecture. Additionally I had
> > a problem with CMA usage on my MIPS machine due to some struct page-related issue. Moving to the
> > memblock allocator fixed the problem and gave us benefits like smaller memory consumption,
> > powerful memblock API to be used within the arch code.
>
> [...]
>
> > @[email protected]. Do you still possess the Octeon MIPS64 platform to test the patchset?
>
> yes, I'd like to test such a serious change on Octeon2. Please include me in the distribution list.
>
Great Alexander! I'll keep you informed then CCing to your email.
> From: Matt Redfearn
> To: 'Serge Semin'
>
> Yes, we are definitely still interested!
>
> We have a bunch of MIPS hardware here which we can test any patches you
> submit with, and should be able to lend a hand to cover as many different
> platforms as we can.
>
> Thanks,
> Matt
>
Hello Matt. It's great! I'll start reworking the patchset asap and keep you
informed when it's finished. Additional notes will be also provided about the
platforms, which for my opinion might not work and couldn't be properly
quick-fixed.
Regards,
-Sergey
Hi Serge,
On 19.12.2017 21:14, Serge Semin wrote:
> Hello folks,
>
> Almost a year ago I sent a patchset to the Linux MIPS community. The main target of the patchset
> was to get rid from the old bootmem allocator usage at the MIPS architecture. Additionally I had
> a problem with CMA usage on my MIPS machine due to some struct page-related issue. Moving to the
> memblock allocator fixed the problem and gave us benefits like smaller memory consumption,
> powerful memblock API to be used within the arch code.
>
> @[email protected]. Could you confirm if you are still interested in the patchset and
> ready to update the Loongson64 platform code so one would be compatible with it?
I am currently working on different projects, but I will try to do my
best to help with this (but just might require more time than I would
otherwise need).
Given the scale of the task and the amount of testing required, I expect
that it will take a while to get a complete set of patches reviewed and
tested which should hopefully be enough to fix support for all platforms
affected.
Marcin
On Fri, Dec 22, 2017 at 08:18:09AM +0100, Marcin Nowakowski <[email protected]> wrote:
> Hi Serge,
>
> On 19.12.2017 21:14, Serge Semin wrote:
> >Hello folks,
> >
> >Almost a year ago I sent a patchset to the Linux MIPS community. The main target of the patchset
> >was to get rid from the old bootmem allocator usage at the MIPS architecture. Additionally I had
> >a problem with CMA usage on my MIPS machine due to some struct page-related issue. Moving to the
> >memblock allocator fixed the problem and gave us benefits like smaller memory consumption,
> >powerful memblock API to be used within the arch code.
> >
> >@[email protected]. Could you confirm if you are still interested in the patchset and
> >ready to update the Loongson64 platform code so one would be compatible with it?
>
> I am currently working on different projects, but I will try to do my best
> to help with this (but just might require more time than I would otherwise
> need).
> Given the scale of the task and the amount of testing required, I expect
> that it will take a while to get a complete set of patches reviewed and
> tested which should hopefully be enough to fix support for all platforms
> affected.
>
Hello Marcin,
Thanks for the willing to help. I'll CC the patchset to you when it's ready.
-Sergey
> Marcin