MIPS Coherent Processor Systems (CPS), which include the MT7621 SoC, support
deep sleep idle states and have a specific cpuidle driver for them.
Enable support for it, while also switching from constant timer ticks to the
idle dynticks model, with the TEO governor.
Run-tested on a Redmi AC2100.
S
In order to fix random hangs on MT7621, we've been selecting
WEAK_REORDERING_BEYOND_LLSC for years [1]. However, these random hangs have been
most likely caused by an oversight in the MIPS implementation of the kernel
memory consistency model, which has already been fixed for some time (and
backpor
We've been carrying this patch for many years [1], in order to fix a timer
calibration issue on MT7621. Turns out, after retesting with a recent kernel
(Linux 5.10), the system works perfectly fine without it (no rcu_sched stalls
or inconsistent BogoMIPS values across CPUs).
Manually refreshed:
32
Drop two obsolete kernel patches and enable support for cpuidle. More details
in each individual patch.
v2: Don't include the generic clock events broadcast patch, as SMP already
selects GENERIC_CLOCKEVENTS_BROADCAST.
Rui Salvaterra (3):
ramips/mt7621: drop the weak reordering patch
ramips/mt
On Tue, Mar 30, 2021 at 10:52:35PM +0100, Rui Salvaterra wrote:
> Hi, Ilya,
>
> On Tue, 30 Mar 2021 at 20:56, Ilya Lipnitskiy
> wrote:
> >
> > "config MIPS_CPS_CPUIDLE" has "select GENERIC_CLOCKEVENTS_BROADCAST if
> > SMP", so it doesn't seem necessary to force it in a dedicated patch.
>
> Oh,
Hi, Ilya,
On Tue, 30 Mar 2021 at 20:56, Ilya Lipnitskiy wrote:
>
> "config MIPS_CPS_CPUIDLE" has "select GENERIC_CLOCKEVENTS_BROADCAST if
> SMP", so it doesn't seem necessary to force it in a dedicated patch.
Oh, indeed?
*goes to look*
Crap, don't know how I missed it [1]. Thanks for the heads
Hi Rui,
On Tue, Mar 30, 2021 at 12:36 PM Rui Salvaterra wrote:
> The GENERIC_CLOCKEVENTS_BROADCAST selection, however, is an unrelated change,
> and must be kept. It's required by cpuidle, which is supported by the
> architecture and will be enabled in the future. Thus, create a new patch
> exclu
MIPS Coherent Processor Systems (CPS), which include the MT7621 SoC, support
deep sleep idle states and have a specific cpuidle driver for them.
Enable support for it, while also switching from constant timer ticks to the
idle dynticks model, with the TEO governor.
Run-tested on a Redmi AC2100.
S
Drop two obsolete kernel patches and enable support for cpuidle. More details
in each individual patch.
Rui Salvaterra (3):
ramips/mt7621: drop the weak reordering patch
ramips/mt7621: drop the timer recalibration patch
ramips/mt7621: enable support for cpuidle
target/linux/ramips/mt7621/c
In order to fix random hangs on MT7621, we've been selecting
WEAK_REORDERING_BEYOND_LLSC for years [1]. However, these random hangs have been
most likely caused by an oversight in the MIPS implementation of the kernel
memory consistency model, which has already been fixed for some time (and
backpor
We've been carrying this patch for many years [1], in order to fix a timer
calibration issue on MT7621. Turns out, after retesting with a recent kernel
(Linux 5.10), the system works perfectly fine without it (no rcu_sched stalls
or inconsistent BogoMIPS values across CPUs).
The GENERIC_CLOCKEVENT
On Tue, 30 Mar 2021 at 19:27, Ilya Lipnitskiy wrote:
>
> My opinion is go ahead and supersede the series with your proposed
> updates, since no maintainer appears to have spent any time on this
> yet.
That as my reasoning too. Will respin, thanks!
___
On 3/30/21 10:53 AM, Paul Spooren wrote:
The ./usign folder is added to every OpenWrt image, it should only
contain the most necessary keys. At this point it contains both a
selection of personal developer keys and keys of EOL releases.
Remove them all and only keep the 21.02 key.
A future comm
On Tue, Mar 30, 2021 at 2:43 AM Rui Salvaterra wrote:
>
> On Tue, 16 Mar 2021 at 09:02, Rui Salvaterra wrote:
> >
> > Although the title is trivial, the potential impact of these changes is
> > enough
> > in itself to warrant more extensive justification, which follows:
> >
> > 202-weak_reorderi
On Tue, 16 Mar 2021 at 09:02, Rui Salvaterra wrote:
>
> Although the title is trivial, the potential impact of these changes is enough
> in itself to warrant more extensive justification, which follows:
>
> 202-weak_reordering.patch - In order to fix random hangs on MT7621, we've been
> selecting
The ./usign folder is added to every OpenWrt image, it should only
contain the most necessary keys. At this point it contains both a
selection of personal developer keys and keys of EOL releases.
Remove them all and only keep the 21.02 key.
A future commit should add a "next release" key, which i
16 matches
Mail list logo