Skip to contentSkip to conversation

x86/timer: Fix boot on Intel systems using ITSSPRC static PIT clock gating

PATCH New Andrew Cooper 1 participant 1 message Jan 27, 2021

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 x86/timer: Fix boot on Intel systems using ITSSPRC static PIT clock gating

Recent Intel client devices have disabled the legacy PIT for powersaving reasons, breaking compatibility with a traditional IBM PC. Xen depends on a legacy timer interrupt to check that the IO-APIC/PIC routing is configured correctly, and fails to boot with:

(XEN) *******************************
(XEN) Panic on CPU 0:
(XEN) IO-APIC + timer doesn't work! Boot with apic_verbosity=debug and send report. Then try booting with the noapic option (XEN) *******************************

While this setting can be undone by Xen, the details of how to differ by chipset, and would be very short sighted for battery based devices. See bit 2 "8254 Static Clock Gating Enable" in:

https://edc.intel.com/content/www/us/en/design/products-and-solutions/processors-and-chipsets/comet-lake-u/intel-400-series-chipset-on-package-platform-controller-hub-register-database/itss-power-reduction-control-itssprc-offset-3300/

All impacted systems have an HPET, but there is no indication of the absence of PIT functionality, nor a suitable way to probe for its absence. As a short term fix, reconfigure the HPET into legacy replacement mode. A better longterm fix would be to avoid the reliance on the timer interrupt entirely.

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Wei Liu <[email protected]>
CC: Aaron Janse <[email protected]>
CC: Jason Andryuk <[email protected]>
CC: Ondrej Balaz <[email protected]>
CC: Tamas K Lengyel <[email protected]>
CC: Marek Marczykowski-Górecki <[email protected]>

v2:
* Fix build - misplaced bracket.
* Tweak description of SETVAL.

Slightly RFC. On older platforms this does generate some spurious PIC interrupts during boot, but they're benign. I was hoping to have time to fix those too, but I'm getting an increasing number of requests to post this patch.

Other followup actions:
* Overhaul our setup logic. Large quantities of it is legacy for pre-64bit systems, and not applicable to Xen these days.
* Have Xen turn the PIT off when it isn't being used as the timesource. Its a waste of time servicing useless interrupts.
* Make clocksource=pit not enter an infinite loop on these systems
* Drop all references to noapic. These days, the only thing it will ever do is make a bad situation worse.

---
 xen/arch/x86/hpet.c | 64 ++++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 63 insertions(+), 1 deletion(-)
xen/arch/x86/hpet.c +63 -1
index e6fab8acd8..876c5e4769 100644
@@ -758,7 +758,7 @@ static u32 *hpet_boot_cfg;
 u64 __init hpet_setup(void)
 {
     static u64 __initdata hpet_rate;
-    u32 hpet_id, hpet_period;
+    unsigned int hpet_id, hpet_period, hpet_cfg;
     unsigned int last, rem;
 
     if ( hpet_rate )
@@ -793,6 +793,68 @@ u64 __init hpet_setup(void)
     if ( (rem * 2) > hpet_period )
         hpet_rate++;
 
+    /*
+     * Intel chipsets from Skylake/ApolloLake onwards can statically clock
+     * gate the 8259 PIT.  This option is enabled by default in slightly later
+     * systems, as turning the PIT off is a prerequisite to entering the C11
+     * power saving state.
+     *
+     * Xen currently depends on the legacy timer interrupt being active while
+     * IRQ routing is configured.
+     *
+     * Reconfigure the HPET into legacy mode to re-establish the timer
+     * interrupt.
+     */
+    if ( hpet_id & HPET_ID_LEGSUP &&
+         !((hpet_cfg = hpet_read32(HPET_CFG)) & HPET_CFG_LEGACY) )
+    {
+        unsigned int c0_cfg, ticks, count;
+
+        /* Stop the main counter. */
+        hpet_write32(hpet_cfg & ~HPET_CFG_ENABLE, HPET_CFG);
+
+        /* Reconfigure channel 0 to be 32bit periodic. */
+        c0_cfg = hpet_read32(HPET_Tn_CFG(0));
+        c0_cfg |= (HPET_TN_ENABLE | HPET_TN_PERIODIC | HPET_TN_SETVAL |
+                   HPET_TN_32BIT);
+        hpet_write32(c0_cfg, HPET_Tn_CFG(0));
+
+        /*
+         * The exact period doesn't have to match a legacy PIT.  All we need
+         * is an interrupt queued up via the IO-APIC to check routing.
+         *
+         * Use HZ as the frequency.
+         */
+        ticks = ((SECONDS(1) / HZ) * div_sc(hpet_rate, SECONDS(1), 32)) >> 32;
+
+        count = hpet_read32(HPET_COUNTER);
+
+        /*
+         * HPET_TN_SETVAL above is atrociously documented in the spec.
+         *
+         * Periodic HPET channels have a main comparator register, and
+         * separate accumulator register.  Each time an interrupt is
+         * generated, the accumulator register is re-added to the comparator
+         * set up the new period.
+         *
+         * Normally, writes to the CMP register update both registers.
+         * However, under these semantics, it is impossible to set up a
+         * periodic timer correctly without the main HPET counter being at 0.
+         *
+         * Instead, HPET_TN_SETVAL is a self-clearing control bit which we can
+         * use for periodic timers to mean that the second write to CMP
+         * updates the accumulator only, and not the absolute comparator
+         * value.
+         *
+         * This lets us set a period when the main counter isn't at 0.
+         */
+        hpet_write32(count + ticks, HPET_Tn_CMP(0));
+        hpet_write32(ticks,         HPET_Tn_CMP(0));
+
+        /* Restart the main counter, and legacy mode. */
+        hpet_write32(hpet_cfg | HPET_CFG_ENABLE | HPET_CFG_LEGACY, HPET_CFG);
+    }
+
     return hpet_rate;
 }
 
Signature · 1 line
2.11.0