lkml.org 
[lkml]   [2018]   [Mar]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [PATCH v2] irqchip/gic-v3: Ensure GICR_CTLR.EnableLPI=0 is observed before enabling
On Sat, 17 Mar 2018 16:11:06 +0000,
Shanker Donthineni wrote:
>
> Hi Marc,
>
> On 03/17/2018 08:34 AM, Marc Zyngier wrote:
> > On Thu, 15 Mar 2018 09:31:27 -0500
> > Shanker Donthineni <shankerd@codeaurora.org> wrote:
> >
> >> The definition of the GICR_CTLR.RWP control bit was expanded to indicate
> >> status of changing GICR_CTLR.EnableLPI from 1 to 0 is being in progress
> >> or completed. Software must observe GICR_CTLR.RWP==0 after clearing
> >> GICR_CTLR.EnableLPI from 1 to 0 and before writing GICR_PENDBASER and/or
> >> GICR_PROPBASER, otherwise behavior is UNPREDICTABLE.
> >>
> >> Signed-off-by: Shanker Donthineni <shankerd@codeaurora.org>
> >> ---
> >> Changes since v1:
> >> -Moved LPI disable code to a seperate function as Marc suggested.
> >> -Mark's suggestion to use readl_relaxed_poll_timeout() helper functions.
> >>
> >> drivers/irqchip/irq-gic-v3-its.c | 66 ++++++++++++++++++++++++++++++--------
> >> include/linux/irqchip/arm-gic-v3.h | 1 +
> >> 2 files changed, 53 insertions(+), 14 deletions(-)
> >>
> >> diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c
> >> index 1d3056f..cba71a7 100644
> >> --- a/drivers/irqchip/irq-gic-v3-its.c
> >> +++ b/drivers/irqchip/irq-gic-v3-its.c
> >> @@ -33,6 +33,8 @@
> >> #include <linux/of_platform.h>
> >> #include <linux/percpu.h>
> >> #include <linux/slab.h>
> >> +#include <linux/time64.h>
> >> +#include <linux/iopoll.h>
> >>
> >> #include <linux/irqchip.h>
> >> #include <linux/irqchip/arm-gic-v3.h>
> >> @@ -1875,16 +1877,6 @@ static void its_cpu_init_lpis(void)
> >> gic_data_rdist()->pend_page = pend_page;
> >> }
> >>
> >> - /* Disable LPIs */
> >> - val = readl_relaxed(rbase + GICR_CTLR);
> >> - val &= ~GICR_CTLR_ENABLE_LPIS;
> >> - writel_relaxed(val, rbase + GICR_CTLR);
> >> -
> >> - /*
> >> - * Make sure any change to the table is observable by the GIC.
> >> - */
> >> - dsb(sy);
> >> -
> >> /* set PROPBASE */
> >> val = (page_to_phys(gic_rdists->prop_page) |
> >> GICR_PROPBASER_InnerShareable |
> >> @@ -3288,13 +3280,59 @@ static bool gic_rdists_supports_plpis(void)
> >> return !!(gic_read_typer(gic_data_rdist_rd_base() + GICR_TYPER) & GICR_TYPER_PLPIS);
> >> }
> >>
> >> +static int redist_disable_lpis(void)
> >> +{
> >> + void __iomem *rbase = gic_data_rdist_rd_base();
> >> + u64 val;
> >> + int ret;
> >> +
> >> + if (!gic_rdists_supports_plpis()) {
> >> + pr_info("CPU%d: LPIs not supported\n", smp_processor_id());
> >> + return -ENXIO;
> >> + }
> >> +
> >> + val = readl_relaxed(rbase + GICR_CTLR);
> >> + if (!(val & GICR_CTLR_ENABLE_LPIS))
> >> + return 0;
> >> +
> >> + /* Disable LPIs */
> >> + val &= ~GICR_CTLR_ENABLE_LPIS;
> >> + writel_relaxed(val, rbase + GICR_CTLR);
> >> +
> >> + /* Make sure any change to GICR_CTLR is observable by the GIC */
> >> + dsb(sy);
> >> +
> >> + /* Wait for GICR_CTLR.GICR_CTLR_ENABLE_LPIS==0 or timeout */
> >> + ret = readl_relaxed_poll_timeout_atomic(rbase + GICR_CTLR, val,
> >> + !(val & GICR_CTLR_ENABLE_LPIS), 1,
> >> + USEC_PER_SEC);
> >
> > Why do you need to poll on EnableLPIs? It is supposed to be immediately
> > written. It seems to me that the sequence should be:
> >
>
> Agree we don't need to poll EnableLPIs here. I'll remove.
>
> > - Check EnableLPis
> > - If set to 1, scream about bootloader being braindead
> > - Write EnableLPIs to 0
> > - Read back EnableLpis:
> > - If still set 1, bail out, we're doomed. Scream about the HW being
> > braindead
> > - Poll RWP
> > - If timeout, bail out, the HW has locked up. Scream again.
> >
> > I think this should cover the whole thing. An alternative would be to
> > place the poll between the write and read-back, but that doesn't change
> > anything.
> >
>
> I'll prefer your second approach. Please comment on the below change before
> I post v3.
>
> static int redist_disable_lpis(void)
> {
> void __iomem *rbase = gic_data_rdist_rd_base();
> u64 val;
> int ret;
>
> if (!gic_rdists_supports_plpis()) {
> pr_info("CPU%d: LPIs not supported\n", smp_processor_id());
> return -ENXIO;
> }
>
> if (!(readl(rbase + GICR_CTLR) & GICR_CTLR_ENABLE_LPIS))
> return 0;
>
> pr_warn("CPU%d: Trying to disable LPIs\n", smp_processor_id());

I'd be much more clear:

pr_warn("CPU%d: Booted with LPIs enabled, memory probably corrupted\n",
smp_processor_id());
add_taint(TAINT_CRAP, LOCKDEP_STILL_OK);

> val &= ~GICR_CTLR_ENABLE_LPIS;
> writel_relaxed(val, rbase + GICR_CTLR);
>
> /* Make sure any change to GICR_CTLR is observable by the GIC */
> dsb(sy);
>
> /**
> * Software must observe RWP==0 after clearing GICR_CTLR.EnableLPIs
> * from 1 to 0 before programming GICR_PEND{PROP}BASER registers.
> * Bail out the driver probe() in case of timeout.
> */
> ret = readl_relaxed_poll_timeout_atomic(rbase + GICR_CTLR, val,
> !(val & GICR_CTLR_RWP), 1,
> USEC_PER_SEC);
> if (ret) {
> pr_err("CPU%d: EnableLPIs not observed\n", smp_processor_id());

Let's call a spade a spade: the redistributor hasn't reacted after a
whole second, it has probably locked-up. Just say so, because your
message doesn't mean much to me.

> return ret;
> }
>
> /**
> * After it has been written to 1, it is IMPLEMENTATION DEFINED whether
> * the bit GICR_CTLR.EnableLPI becomes RES1 or can be cleared to 0.
> * Bail out the driver probe() on systems where it's RES1.
> */
> if (readl_relaxed(rbase + GICR_CTLR) & GICR_CTLR_ENABLE_LPIS) {
> pr_err("CPU%d: Failed to disable LPIs\n", smp_processor_id());
> return -EPERM;

EPERM? EBUSY was a better approximation...

Thanks,

M.

--
Jazz is not dead, it just smell funny.

\
 
 \ /
  Last update: 2018-03-17 18:19    [W:0.090 / U:1.372 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site