Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1409651

[PATCH 4.4 28/86] irqchip/gic: Ensure ordering between read of INTACK and shared data

From Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Newsgroups linux.kernel
Subject [PATCH 4.4 28/86] irqchip/gic: Ensure ordering between read of INTACK and shared data
Date 2016-05-31 00:10 +0200
Message-ID <rEDxN-1RI-39@gated-at.bofh.it> (permalink)
References <rECs1-VP-3@gated-at.bofh.it>
Organization linux.* mail to news gateway

Show all headers | View raw

4.4-stable review patch.  If anyone has any objections, please let me know.

------------------

From: Will Deacon <will.deacon@arm.com>

commit f86c4fbd930ff6fecf3d8a1c313182bd0f49f496 upstream.

When an IPI is generated by a CPU, the pattern looks roughly like:

  <write shared data>
  smp_wmb();
  <write to GIC to signal SGI>

On the receiving CPU we rely on the fact that, once we've taken the
interrupt, then the freshly written shared data must be visible to us.
Put another way, the CPU isn't going to speculate taking an interrupt.

Unfortunately, this assumption turns out to be broken.

Consider that CPUx wants to send an IPI to CPUy, which will cause CPUy
to read some shared_data. Before CPUx has done anything, a random
peripheral raises an IRQ to the GIC and the IRQ line on CPUy is raised.
CPUy then takes the IRQ and starts executing the entry code, heading
towards gic_handle_irq. Furthermore, let's assume that a bunch of the
previous interrupts handled by CPUy were SGIs, so the branch predictor
kicks in and speculates that irqnr will be <16 and we're likely to
head into handle_IPI. The prefetcher then grabs a speculative copy of
shared_data which contains a stale value.

Meanwhile, CPUx gets round to updating shared_data and asking the GIC
to send an SGI to CPUy. Internally, the GIC decides that the SGI is
more important than the peripheral interrupt (which hasn't yet been
ACKed) but doesn't need to do anything to CPUy, because the IRQ line
is already raised.

CPUy then reads the ACK register on the GIC, sees the SGI value which
confirms the branch prediction and we end up with a stale shared_data
value.

This patch fixes the problem by adding an smp_rmb() to the IPI entry
code in gic_handle_irq. As it turns out, the combination of a control
dependency and an ISB instruction from the EOI in the GICv3 driver is
enough to provide the ordering we need, so we add a comment there
justifying the absence of an explicit smp_rmb().

Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

---
 drivers/irqchip/irq-gic-v3.c |    7 +++++++
 drivers/irqchip/irq-gic.c    |    8 ++++++++
 2 files changed, 15 insertions(+)

--- a/drivers/irqchip/irq-gic-v3.c
+++ b/drivers/irqchip/irq-gic-v3.c
@@ -361,6 +361,13 @@ static asmlinkage void __exception_irq_e
 			if (static_key_true(&supports_deactivate))
 				gic_write_dir(irqnr);
 #ifdef CONFIG_SMP
+			/*
+			 * Unlike GICv2, we don't need an smp_rmb() here.
+			 * The control dependency from gic_read_iar to
+			 * the ISB in gic_write_eoir is enough to ensure
+			 * that any shared data read by handle_IPI will
+			 * be read after the ACK.
+			 */
 			handle_IPI(irqnr, regs);
 #else
 			WARN_ONCE(true, "Unexpected SGI received!\n");
--- a/drivers/irqchip/irq-gic.c
+++ b/drivers/irqchip/irq-gic.c
@@ -347,6 +347,14 @@ static void __exception_irq_entry gic_ha
 			if (static_key_true(&supports_deactivate))
 				writel_relaxed(irqstat, cpu_base + GIC_CPU_DEACTIVATE);
 #ifdef CONFIG_SMP
+			/*
+			 * Ensure any shared data written by the CPU sending
+			 * the IPI is read after we've read the ACK register
+			 * on the GIC.
+			 *
+			 * Pairs with the write barrier in gic_raise_softirq
+			 */
+			smp_rmb();
 			handle_IPI(irqnr, regs);
 #endif
 			continue;

Back to linux.kernel | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

[PATCH 4.4 00/86] 4.4.12-stable review Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:00 +0200
  [PATCH 4.4 83/86] hpfs: implement the show_options method Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:00 +0200
  [PATCH 4.4 80/86] SIGNAL: Move generic copy_siginfo() to signal.h Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:00 +0200
  [PATCH 4.4 70/86] serial: samsung: Reorder the sequence of clock control when call s3c24xx_serial_set_termios() Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:00 +0200
  [PATCH 4.4 63/86] tty: vt, return error when con_startup fails Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:00 +0200
  [PATCH 4.4 54/86] USB: leave LPM alone if possible when binding/unbinding interface drivers Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:00 +0200
  [PATCH 4.4 75/86] ALSA: hda - Fix headphone noise on Dell XPS 13 9360 Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:00 +0200
  [PATCH 4.4 73/86] mcb: Fixed bar number assignment for the gdd Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:00 +0200
  [PATCH 4.4 77/86] ALSA: hda - Fix headset mic detection problem for one Dell machine Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:00 +0200
  [PATCH 4.4 78/86] IB/srp: Fix a debug kernel crash Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:00 +0200
  [PATCH 4.4 01/86] Btrfs: dont use src fd for printk Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:10 +0200
  [PATCH 4.4 49/86] USB: serial: option: add more ZTE device ids Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:10 +0200
  [PATCH 4.4 86/86] kbuild: move -Wunused-const-variable to W=1 warning level Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:20 +0200
  [PATCH 4.4 55/86] usb: gadget: udc: core: Fix argument of dev_err() in usb_gadget_map_request() Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-30 23:50 +0200
  [PATCH 4.4 02/86] perf/x86/intel/pt: Generate PMI in the STOP region as well Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:00 +0200
  [PATCH 4.4 47/86] USB: serial: io_edgeport: fix memory leaks in probe error path Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:00 +0200
  [PATCH 4.4 11/86] arm/arm64: KVM: Enforce Break-Before-Make on Stage-2 page tables Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:00 +0200
  [PATCH 4.4 43/86] USB: serial: mxuport: fix use-after-free in probe error path Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:00 +0200
  [PATCH 4.4 42/86] mei: bus: call mei_cl_read_start under device lock Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:00 +0200
  [PATCH 4.4 46/86] USB: serial: io_edgeport: fix memory leaks in attach error path Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:00 +0200
  [PATCH 4.4 51/86] usb: gadget: f_fs: Fix EFAULT generation for async read operations Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:00 +0200
  [PATCH 4.4 45/86] USB: serial: quatech2: fix use-after-free in probe error path Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:00 +0200
  [PATCH 4.4 41/86] mei: amthif: discard not read messages Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:00 +0200
  [PATCH 4.4 31/86] mmc: mmc: Fix partition switch timeout for some eMMCs Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 15/86] fs/cifs: correctly to anonymous authentication for the LANMAN authentication Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 29/86] irqchip/gic-v3: Configure all interrupts as non-secure Group-1 Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 06/86] arm64: Fix typo in the pmdp_huge_get_and_clear() definition Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 13/86] remove directory incorrectly tries to set delete on close on non-empty directories Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 32/86] mmc: sdhci-acpi: Remove MMC_CAP_BUS_WIDTH_TEST for Intel controllers Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 19/86] ring-buffer: Use long for nr_pages to avoid overflow failures Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 03/86] perf/core: Fix perf_event_open() vs. execve() race Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 20/86] ring-buffer: Prevent overflow of size in ring_buffer_resize() Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 28/86] irqchip/gic: Ensure ordering between read of INTACK and shared data Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 05/86] ext4: iterate over buffer heads correctly in move_extent_per_page() Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 12/86] kvm: arm64: Fix EC field in inject_abt64 Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 09/86] arm64: Implement pmdp_set_access_flags() for hardware AF/DBM Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 07/86] arm64: Ensure pmd_present() returns false after pmd_mknotpresent() Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 36/86] mmc: sdhci-pci: Remove MMC_CAP_BUS_WIDTH_TEST for Intel controllers Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 16/86] fs/cifs: correctly to anonymous authentication for the NTLM(v1) authentication Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 33/86] ACPI / osi: Fix an issue that acpi_osi=!* cannot disable ACPICA internal strings Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 18/86] asix: Fix offset calculation in asix_rx_fixup() causing slow transmissions Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 23/86] crypto: sun4i-ss - Replace spinlock_bh by spin_lock_irq{save|restore} Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 38/86] Bluetooth: vhci: purge unhandled skbs Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 27/86] Input: pwm-beeper - fix - scheduling while atomic Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 30/86] can: fix handling of unmodifiable configuration options Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 17/86] fs/cifs: correctly to anonymous authentication for the NTLM(v2) authentication Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 10/86] arm64: cpuinfo: Missing NULL terminator in compat_hwcap_str Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 24/86] clk: qcom: msm8916: Fix crypto clock flags Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 14/86] fs/cifs: correctly to anonymous authentication via NTLMSSP Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 22/86] crypto: talitos - fix ahash algorithms registration Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 08/86] arm64: Implement ptep_set_access_flags() for hardware AF/DBM Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  [PATCH 4.4 26/86] mfd: omap-usb-tll: Fix scheduling while atomic BUG Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2016-05-31 00:10 +0200
  Re: [PATCH 4.4 00/86] 4.4.12-stable review Shuah Khan <shuahkh@osg.samsung.com> - 2016-06-01 16:30 +0200

csiph-web