| Message ID | 1455140107-3328-1-git-send-email-tony@atomide.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi, On 02/10/2016 03:35 PM, Tony Lindgren wrote: > With CONFIG_DEBUG_RODATA enabled I started noticing imprecise external > aborts on a dm3730 when hitting off idle. These don't seem to happen > on 34xx. > > Pretty much changing anything in the code made these go away, like > changing .config options. At first I though it might be an alignment > issue in the 36xx specific assembly code in sleep34xx.S, or something > related to the recent rodata fixes. But that does not seem to be > the case. It seems to be a timing issue instead. > > Adding few extra nop instructions after the wfi seems to fix the > issue. When adding 5 nops, the errors showed up less often. With > add ed 6 nops, I don't seem to get them at all any longer. > > Cc: Grygorii Strashko <grygorii.strashko@ti.com> > Cc: Nishanth Menon <nm@ti.com> > Cc: Richard Woodruff <r-woodruff2@ti.com> > Cc: Tero Kristo <t-kristo@ti.com> > Signed-off-by: Tony Lindgren <tony@atomide.com> > --- > > Anybody else seen this issue before? I have a series to convert omap3 PM code to using generic SRAM driver but when testing I see an external abort on BBXM off mode resume very similar to this that seems to be timing related caused by using generic SRAM driver to re-copy PM code rather than omap3_sram_restore_context. By tracing the resume path I believe the abort is caused by omap3_intc_resume_idle in pm34xx.c, which writes to two registers in the INTC. Removing this call makes the abort unreproducible in my experiments and changing the writes to reads causes a bus lock, so I'm pretty confident it's coming from this call attempting to write to an idled INTC. Advisory 1.106 in DM3730 Silicon Errata Document [1] describes an issue with "MPU Leaves MSTANDBY State Before IDLEREQ of Interrupt Controller is Released" which sounds like a very similar failure situation to what we are seeing even though the timing of INTC access is a bit further from WFI exit than it describes. Perhaps there are more conditions where it can occur. Pushed my WIP branch for SRAM series that shows the same failure here [2]. Regards, Dave [1] http://www.ti.com/lit/er/sprz319f/sprz319f.pdf [2] https://github.com/dgerlach/linux-pm/tree/beagle-sram-wip-v4.5 > > --- > arch/arm/mach-omap2/sleep34xx.S | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/arch/arm/mach-omap2/sleep34xx.S b/arch/arm/mach-omap2/sleep34xx.S > index 1b9f052..0fbaa08 100644 > --- a/arch/arm/mach-omap2/sleep34xx.S > +++ b/arch/arm/mach-omap2/sleep34xx.S > @@ -264,6 +264,12 @@ ENTRY(omap3_do_wfi) > nop > nop > nop > + nop > + nop > + nop > + nop > + nop > + nop > > /* > * This function implements the erratum ID i581 WA: > -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/arch/arm/mach-omap2/sleep34xx.S b/arch/arm/mach-omap2/sleep34xx.S index 1b9f052..0fbaa08 100644 --- a/arch/arm/mach-omap2/sleep34xx.S +++ b/arch/arm/mach-omap2/sleep34xx.S @@ -264,6 +264,12 @@ ENTRY(omap3_do_wfi) nop nop nop + nop + nop + nop + nop + nop + nop /* * This function implements the erratum ID i581 WA:
With CONFIG_DEBUG_RODATA enabled I started noticing imprecise external aborts on a dm3730 when hitting off idle. These don't seem to happen on 34xx. Pretty much changing anything in the code made these go away, like changing .config options. At first I though it might be an alignment issue in the 36xx specific assembly code in sleep34xx.S, or something related to the recent rodata fixes. But that does not seem to be the case. It seems to be a timing issue instead. Adding few extra nop instructions after the wfi seems to fix the issue. When adding 5 nops, the errors showed up less often. With add ed 6 nops, I don't seem to get them at all any longer. Cc: Grygorii Strashko <grygorii.strashko@ti.com> Cc: Nishanth Menon <nm@ti.com> Cc: Richard Woodruff <r-woodruff2@ti.com> Cc: Tero Kristo <t-kristo@ti.com> Signed-off-by: Tony Lindgren <tony@atomide.com> --- Anybody else seen this issue before? --- arch/arm/mach-omap2/sleep34xx.S | 6 ++++++ 1 file changed, 6 insertions(+)