Re: bcm2837: tons of DMA mask not set in 4.18.0-rc6-next-20180727
Hi Geert,
On 21/08/18 16:45, Geert Uytterhoeven wrote:
Hi Robin,
CC Lee (mfd), Wolfram (i2c), Marek (bd9571wmv + recent da9063 hacking)
On Fri, Jul 27, 2018 at 6:28 PM Robin Murphy wrote:
On 27/07/18 17:17, Stefan Wahren wrote:
i was noticed that recent linux-next on Raspberry Pi 3 (32 bit) doesn't boot
anymore. So i tested todays linux-next with multi_v7_defconfig and i'm getting
a lot of DMA mask not set warnings. At least USB dwc2 is broken.
Any suggestions before i start to bisect?
https://patchwork.kernel.org/patch/10547231/
Although the upside of this hiccup is that it should give us a fair view
of how many active drivers are still relying on implicit default DMA
masks - I'm now waiting for kernelci's boot chart to catch up with
today's -next to see the full fallout...
I'm still seeing the following on r8a7791/koelsch with a DA9063 PMIC, which
is an MFD device:
da9063-watchdog da9063-watchdog: DMA mask not set
da9063-rtc da9063-rtc: DMA mask not set
The dma masks (both normal and coherent) are NULL, and inherited from the
i2c parent device (6-0058):
pmic@58 {
compatible = "dlg,da9063";
reg = <0x58>;
interrupt-parent = <&irqc0>;
interrupts = <2 IRQ_TYPE_LEVEL_LOW>;
interrupt-controller;
rtc {
compatible = "dlg,da9063-rtc";
};
wdt {
compatible = "dlg,da9063-watchdog";
};
};
The DA9063 MFD driver creates more subdevices than the two listed above,
but the warning doesn't trigger for them, presumably because they don't
have corresponding nodes in DT.
On r8a779{5,6,65}*/salvator-x, there's also an MFD device (BD9571WMC PMIC)
on an i2c bus, inheriting NULL DMA masks from its parent, but it doesn't
have any subnodes in DT, and the warning isn't triggered neither.
I think the question here is why I2C slave devices are getting the
chance to be configured as DMA masters at all :/
Are they getting added to the wrong bus somehow?
Robin.
Anyone with a clue?
Thanks!
Gr{oetje,eeting}s,
Geert
Re: bcm2837: tons of DMA mask not set in 4.18.0-rc6-next-20180727
On Tue, Aug 21, 2018 at 05:45:43PM +0200, Geert Uytterhoeven wrote: > Hi Robin, > > CC Lee (mfd), Wolfram (i2c), Marek (bd9571wmv + recent da9063 hacking) > > On Fri, Jul 27, 2018 at 6:28 PM Robin Murphy wrote: > > On 27/07/18 17:17, Stefan Wahren wrote: > > > i was noticed that recent linux-next on Raspberry Pi 3 (32 bit) doesn't > > > boot anymore. So i tested todays linux-next with multi_v7_defconfig and > > > i'm getting a lot of DMA mask not set warnings. At least USB dwc2 is > > > broken. > > > > > > Any suggestions before i start to bisect? > > > > https://patchwork.kernel.org/patch/10547231/ > > > > Although the upside of this hiccup is that it should give us a fair view > > of how many active drivers are still relying on implicit default DMA > > masks - I'm now waiting for kernelci's boot chart to catch up with > > today's -next to see the full fallout... > > I'm still seeing the following on r8a7791/koelsch with a DA9063 PMIC, which > is an MFD device: > > da9063-watchdog da9063-watchdog: DMA mask not set > da9063-rtc da9063-rtc: DMA mask not set > > The dma masks (both normal and coherent) are NULL, and inherited from the > i2c parent device (6-0058): I think the larger question is: why is the kernel complaining about an unset DMA mask for a device that does not have any capability to perform DMA? Surely the warning is wrong. -- RMK's Patch system: http://www.armlinux.org.uk/developer/patches/ FTTC broadband for 0.8mile line in suburbia: sync at 13.8Mbps down 630kbps up According to speedtest.net: 13Mbps down 490kbps up
Re: bcm2837: tons of DMA mask not set in 4.18.0-rc6-next-20180727
Hi Robin,
CC Lee (mfd), Wolfram (i2c), Marek (bd9571wmv + recent da9063 hacking)
On Fri, Jul 27, 2018 at 6:28 PM Robin Murphy wrote:
> On 27/07/18 17:17, Stefan Wahren wrote:
> > i was noticed that recent linux-next on Raspberry Pi 3 (32 bit) doesn't
> > boot anymore. So i tested todays linux-next with multi_v7_defconfig and i'm
> > getting a lot of DMA mask not set warnings. At least USB dwc2 is broken.
> >
> > Any suggestions before i start to bisect?
>
> https://patchwork.kernel.org/patch/10547231/
>
> Although the upside of this hiccup is that it should give us a fair view
> of how many active drivers are still relying on implicit default DMA
> masks - I'm now waiting for kernelci's boot chart to catch up with
> today's -next to see the full fallout...
I'm still seeing the following on r8a7791/koelsch with a DA9063 PMIC, which
is an MFD device:
da9063-watchdog da9063-watchdog: DMA mask not set
da9063-rtc da9063-rtc: DMA mask not set
The dma masks (both normal and coherent) are NULL, and inherited from the
i2c parent device (6-0058):
pmic@58 {
compatible = "dlg,da9063";
reg = <0x58>;
interrupt-parent = <&irqc0>;
interrupts = <2 IRQ_TYPE_LEVEL_LOW>;
interrupt-controller;
rtc {
compatible = "dlg,da9063-rtc";
};
wdt {
compatible = "dlg,da9063-watchdog";
};
};
The DA9063 MFD driver creates more subdevices than the two listed above,
but the warning doesn't trigger for them, presumably because they don't
have corresponding nodes in DT.
On r8a779{5,6,65}*/salvator-x, there's also an MFD device (BD9571WMC PMIC)
on an i2c bus, inheriting NULL DMA masks from its parent, but it doesn't
have any subnodes in DT, and the warning isn't triggered neither.
Anyone with a clue?
Thanks!
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@linux-m68k.org
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
