[D][PATCH v2 0/3] Add kernel parameter 'pci=clearmsi' to clear MSI(X)s early on boot

classic Classic list List threaded Threaded
7 messages Options
Reply | Threaded
Open this post in threaded view
|

[D][PATCH v2 0/3] Add kernel parameter 'pci=clearmsi' to clear MSI(X)s early on boot

Mauricio Faria de Oliveira-3
BugLink: https://bugs.launchpad.net/bugs/1797990

(Note: the patch set for Cosmic/Disco should be identical,
 as both trees' master-next branch is at the 4.18.0-11.12
 level, with some more patches in Cosmic, but doing Disco
 as well for process guidelines/completeness sake.)

[Changelog]
 * v2:
   - Reorder patch 1 as 3 to allow for the next change:
   - Gate the bus-scan differences with the cmdline option
     (patch 3 only).  Now all functional changes are gated.

[Impact]

 * A kexec/crash kernel might get stuck and fail to boot
   (for crash kernel, kdump fails to collect a crashdump)
   if a PCI device is buggy/stuck/looping and triggers a
   continuous flood of MSI(X) interrupts (that the kernel
   does not yet know about).

 * This fix allowed to obtain crashdumps when debugging a
   heavy-load scenario, in which a (heavy-loaded) network
   adapter wouldn't stop triggering MSI-X interrupts ever
   after panic()->kdump kicked in.

 * This fix disables MSI(X) in all PCI devices on early
   boot (this is OK as it's (re-)enabled normally later)
   with a kernel cmdline parameter (disabled by default).

[Test Case]

 * A synthetic test-case is not yet available, however,
   this particular system/workload triggered the problem
   consistently, and it was used for development/testing.

 * We'll update this bug once a synthetic test-case is
   available; we're working on patching QEMU for this.

 * $ dmesg | grep 'Clearing MSI'
   [    0.000000] Clearing MSI/MSI-X enable bits early in boot (quirk)

 * The comparison of 'dmesg -t | sort' has been reviewed
   between option disabled/enabled on boot & kexec modes,
   and only expected differences found (MHz, PIDs, MIPS).

[Regression Potential]

 * The potential area for regressions is early boot,
   particularly effects of applying quirks during PCI
   bus scan, which is changed/broader w/ these patches.

 * However, all quirks are applied based on PCI ID
   matching, so would only apply if actually targeting
   a new device.

 * Moreover, the new quirk is only applied based on
   a kernel cmdline parameter that is disabled by
   default, which constraints even more when this
   is actually in effect.

[Other Info]
 
 * The patch series is still under review/discussion
   upstream, but it's relatively important for Ubuntu
   users at this point, and after internal discussions
   we decided to submit it for SRU.

 * These are links to the linux-pci archive with the
   patches [1, 2, 3]

   [1] [PATCH 1/3] x86/quirks: Scan all busses for early PCI quirks
       https://lore.kernel.org/linux-pci/20181018183721.27467-1-gpiccoli@.../

   [2] [PATCH 2/3] x86/PCI: Export find_cap() to be used in early PCI code
       https://lore.kernel.org/linux-pci/20181018183721.27467-2-gpiccoli@.../

   [3] [PATCH 3/3] x86/quirks: Add parameter to clear MSIs early on boot
       https://lore.kernel.org/linux-pci/20181018183721.27467-3-gpiccoli@.../

Guilherme G. Piccoli (3):
  UBUNTU: SAUCE: x86/PCI: Export find_cap() to be used in early PCI code
  UBUNTU: SAUCE: x86/quirks: Add parameter to clear MSIs early on boot
  UBUNTU: SAUCE: x86/quirks: Scan all busses for early PCI quirks

 .../admin-guide/kernel-parameters.txt         |  6 +++
 arch/x86/include/asm/pci-direct.h             |  2 +
 arch/x86/kernel/aperture_64.c                 | 30 +-------------
 arch/x86/kernel/early-quirks.c                | 41 +++++++++++++++++++
 arch/x86/pci/common.c                         |  4 ++
 arch/x86/pci/early.c                          | 25 +++++++++++
 6 files changed, 80 insertions(+), 28 deletions(-)

--
2.17.1


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

[D][PATCH v2 1/3] UBUNTU: SAUCE: x86/PCI: Export find_cap() to be used in early PCI code

Mauricio Faria de Oliveira-3
From: "Guilherme G. Piccoli" <[hidden email]>

BugLink: https://bugs.launchpad.net/bugs/1797990

This patch exports (and renames) the function find_cap() to be used
in the early PCI quirk code, by the next patch.

This is being moved out from AGP code to generic early-PCI code
since it's not AGP-specific and can be used for any PCI device.
No functional changes intended.

Signed-off-by: Guilherme G. Piccoli <[hidden email]>
Signed-off-by: Mauricio Faria de Oliveira <[hidden email]>
---
 arch/x86/include/asm/pci-direct.h |  1 +
 arch/x86/kernel/aperture_64.c     | 30 ++----------------------------
 arch/x86/pci/early.c              | 25 +++++++++++++++++++++++++
 3 files changed, 28 insertions(+), 28 deletions(-)

diff --git a/arch/x86/include/asm/pci-direct.h b/arch/x86/include/asm/pci-direct.h
index e1084f71a295..3f5b5158bbd7 100644
--- a/arch/x86/include/asm/pci-direct.h
+++ b/arch/x86/include/asm/pci-direct.h
@@ -10,6 +10,7 @@
 extern u32 read_pci_config(u8 bus, u8 slot, u8 func, u8 offset);
 extern u8 read_pci_config_byte(u8 bus, u8 slot, u8 func, u8 offset);
 extern u16 read_pci_config_16(u8 bus, u8 slot, u8 func, u8 offset);
+extern u32 pci_early_find_cap(int bus, int slot, int func, int cap);
 extern void write_pci_config(u8 bus, u8 slot, u8 func, u8 offset, u32 val);
 extern void write_pci_config_byte(u8 bus, u8 slot, u8 func, u8 offset, u8 val);
 extern void write_pci_config_16(u8 bus, u8 slot, u8 func, u8 offset, u16 val);
diff --git a/arch/x86/kernel/aperture_64.c b/arch/x86/kernel/aperture_64.c
index 2c4d5ece7456..365fcc37b2a2 100644
--- a/arch/x86/kernel/aperture_64.c
+++ b/arch/x86/kernel/aperture_64.c
@@ -120,32 +120,6 @@ static u32 __init allocate_aperture(void)
 }
 
 
-/* Find a PCI capability */
-static u32 __init find_cap(int bus, int slot, int func, int cap)
-{
- int bytes;
- u8 pos;
-
- if (!(read_pci_config_16(bus, slot, func, PCI_STATUS) &
- PCI_STATUS_CAP_LIST))
- return 0;
-
- pos = read_pci_config_byte(bus, slot, func, PCI_CAPABILITY_LIST);
- for (bytes = 0; bytes < 48 && pos >= 0x40; bytes++) {
- u8 id;
-
- pos &= ~3;
- id = read_pci_config_byte(bus, slot, func, pos+PCI_CAP_LIST_ID);
- if (id == 0xff)
- break;
- if (id == cap)
- return pos;
- pos = read_pci_config_byte(bus, slot, func,
- pos+PCI_CAP_LIST_NEXT);
- }
- return 0;
-}
-
 /* Read a standard AGPv3 bridge header */
 static u32 __init read_agp(int bus, int slot, int func, int cap, u32 *order)
 {
@@ -234,8 +208,8 @@ static u32 __init search_agp_bridge(u32 *order, int *valid_agp)
  case PCI_CLASS_BRIDGE_HOST:
  case PCI_CLASS_BRIDGE_OTHER: /* needed? */
  /* AGP bridge? */
- cap = find_cap(bus, slot, func,
- PCI_CAP_ID_AGP);
+ cap = pci_early_find_cap(bus, slot,
+ func, PCI_CAP_ID_AGP);
  if (!cap)
  break;
  *valid_agp = 1;
diff --git a/arch/x86/pci/early.c b/arch/x86/pci/early.c
index e5f753cbb1c3..c082b530ee2d 100644
--- a/arch/x86/pci/early.c
+++ b/arch/x86/pci/early.c
@@ -51,6 +51,31 @@ void write_pci_config_16(u8 bus, u8 slot, u8 func, u8 offset, u16 val)
  outw(val, 0xcfc + (offset&2));
 }
 
+u32 pci_early_find_cap(int bus, int slot, int func, int cap)
+{
+ int bytes;
+ u8 pos;
+
+ if (!(read_pci_config_16(bus, slot, func, PCI_STATUS) &
+ PCI_STATUS_CAP_LIST))
+ return 0;
+
+ pos = read_pci_config_byte(bus, slot, func, PCI_CAPABILITY_LIST);
+ for (bytes = 0; bytes < 48 && pos >= 0x40; bytes++) {
+ u8 id;
+
+ pos &= ~3;
+ id = read_pci_config_byte(bus, slot, func, pos+PCI_CAP_LIST_ID);
+ if (id == 0xff)
+ break;
+ if (id == cap)
+ return pos;
+ pos = read_pci_config_byte(bus, slot, func,
+ pos+PCI_CAP_LIST_NEXT);
+ }
+ return 0;
+}
+
 int early_pci_allowed(void)
 {
  return (pci_probe & (PCI_PROBE_CONF1|PCI_PROBE_NOEARLY)) ==
--
2.17.1


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

[D][PATCH v2 2/3] UBUNTU: SAUCE: x86/quirks: Add parameter to clear MSIs early on boot

Mauricio Faria de Oliveira-3
In reply to this post by Mauricio Faria de Oliveira-3
From: "Guilherme G. Piccoli" <[hidden email]>

BugLink: https://bugs.launchpad.net/bugs/1797990

We observed a kdump failure in x86 that was narrowed down to MSI irq
storm coming from a PCI network device. The bug manifests as a lack of
progress in the boot process of kdump kernel, and a flood of kernel
messages like:

[...]
[ 342.265294] do_IRQ: 0.155 No irq handler for vector
[ 342.266916] do_IRQ: 0.155 No irq handler for vector
[ 347.258422] do_IRQ: 14053260 callbacks suppressed
[...]

The root cause of the issue is that kexec process of the kdump kernel
doesn't ensure PCI devices are reset or MSI capabilities are disabled,
so a PCI adapter could produce a huge amount of irqs which would steal
all the processing time for the CPU (specially since we usually restrict
kdump kernel to use a single CPU only).

This patch implements the kernel parameter "pci=clearmsi" to clear the
MSI/MSI-X enable bits in the Message Control register for all PCI devices
during early boot time, thus preventing potential issues in the kexec'ed
kernel. PCI spec also supports/enforces this need (see PCI Local Bus
spec sections 6.8.1.3 and 6.8.2.3).

Suggested-by: Dan Streetman <[hidden email]>
Suggested-by: Gavin Shan <[hidden email]>
Signed-off-by: Guilherme G. Piccoli <[hidden email]>
[mfo: backport to ubuntu-disco:
 - update context lines in pci-direct.h]
Signed-off-by: Mauricio Faria de Oliveira <[hidden email]>
---
 .../admin-guide/kernel-parameters.txt         |  6 ++++
 arch/x86/include/asm/pci-direct.h             |  1 +
 arch/x86/kernel/early-quirks.c                | 32 +++++++++++++++++++
 arch/x86/pci/common.c                         |  4 +++
 4 files changed, 43 insertions(+)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 273d6f515315..9587a2f1137e 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -3110,6 +3110,12 @@
  nomsi [MSI] If the PCI_MSI kernel config parameter is
  enabled, this kernel boot option can be used to
  disable the use of MSI interrupts system-wide.
+ clearmsi [X86] Clears MSI/MSI-X enable bits early in boot
+ time in order to avoid issues like adapters
+ screaming irqs and preventing boot progress.
+ Also, it enforces the PCI Local Bus spec
+ rule that those bits should be 0 in system reset
+ events (useful for kexec/kdump cases).
  noioapicquirk [APIC] Disable all boot interrupt quirks.
  Safety option to keep boot IRQs enabled. This
  should never be necessary.
diff --git a/arch/x86/include/asm/pci-direct.h b/arch/x86/include/asm/pci-direct.h
index 3f5b5158bbd7..14d4aa4e1aed 100644
--- a/arch/x86/include/asm/pci-direct.h
+++ b/arch/x86/include/asm/pci-direct.h
@@ -15,6 +15,7 @@ extern void write_pci_config(u8 bus, u8 slot, u8 func, u8 offset, u32 val);
 extern void write_pci_config_byte(u8 bus, u8 slot, u8 func, u8 offset, u8 val);
 extern void write_pci_config_16(u8 bus, u8 slot, u8 func, u8 offset, u16 val);
 
+extern unsigned int pci_early_clear_msi;
 extern int early_pci_allowed(void);
 
 extern unsigned int pci_early_dump_regs;
diff --git a/arch/x86/kernel/early-quirks.c b/arch/x86/kernel/early-quirks.c
index 50d5848bf22e..e7561314eb5d 100644
--- a/arch/x86/kernel/early-quirks.c
+++ b/arch/x86/kernel/early-quirks.c
@@ -28,6 +28,37 @@
 #include <asm/irq_remapping.h>
 #include <asm/early_ioremap.h>
 
+static void __init early_pci_clear_msi(int bus, int slot, int func)
+{
+ int pos;
+ u16 ctrl;
+
+ if (likely(!pci_early_clear_msi))
+ return;
+
+ pr_info_once("Clearing MSI/MSI-X enable bits early in boot (quirk)\n");
+
+ pos = pci_early_find_cap(bus, slot, func, PCI_CAP_ID_MSI);
+ if (pos) {
+ ctrl = read_pci_config_16(bus, slot, func, pos + PCI_MSI_FLAGS);
+ ctrl &= ~PCI_MSI_FLAGS_ENABLE;
+ write_pci_config_16(bus, slot, func, pos + PCI_MSI_FLAGS, ctrl);
+
+ /* Read again to flush previous write */
+ ctrl = read_pci_config_16(bus, slot, func, pos + PCI_MSI_FLAGS);
+ }
+
+ pos = pci_early_find_cap(bus, slot, func, PCI_CAP_ID_MSIX);
+ if (pos) {
+ ctrl = read_pci_config_16(bus, slot, func, pos + PCI_MSIX_FLAGS);
+ ctrl &= ~PCI_MSIX_FLAGS_ENABLE;
+ write_pci_config_16(bus, slot, func, pos + PCI_MSIX_FLAGS, ctrl);
+
+ /* Read again to flush previous write */
+ ctrl = read_pci_config_16(bus, slot, func, pos + PCI_MSIX_FLAGS);
+ }
+}
+
 static void __init fix_hypertransport_config(int num, int slot, int func)
 {
  u32 htcfg;
@@ -709,6 +740,7 @@ static struct chipset early_qrk[] __initdata = {
  PCI_CLASS_BRIDGE_HOST, PCI_ANY_ID, 0, force_disable_hpet},
  { PCI_VENDOR_ID_BROADCOM, 0x4331,
   PCI_CLASS_NETWORK_OTHER, PCI_ANY_ID, 0, apple_airport_reset},
+ { PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0, early_pci_clear_msi},
  {}
 };
 
diff --git a/arch/x86/pci/common.c b/arch/x86/pci/common.c
index 563049c483a1..94b156f85573 100644
--- a/arch/x86/pci/common.c
+++ b/arch/x86/pci/common.c
@@ -33,6 +33,7 @@ int noioapicreroute = 1;
 #endif
 int pcibios_last_bus = -1;
 unsigned long pirq_table_addr;
+unsigned int pci_early_clear_msi;
 const struct pci_raw_ops *__read_mostly raw_pci_ops;
 const struct pci_raw_ops *__read_mostly raw_pci_ext_ops;
 
@@ -608,6 +609,9 @@ char *__init pcibios_setup(char *str)
  } else if (!strcmp(str, "skip_isa_align")) {
  pci_probe |= PCI_CAN_SKIP_ISA_ALIGN;
  return NULL;
+ } else if (!strcmp(str, "clearmsi")) {
+ pci_early_clear_msi = 1;
+ return NULL;
  } else if (!strcmp(str, "noioapicquirk")) {
  noioapicquirk = 1;
  return NULL;
--
2.17.1


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

[D][PATCH v2 3/3] UBUNTU: SAUCE: x86/quirks: Scan all busses for early PCI quirks

Mauricio Faria de Oliveira-3
In reply to this post by Mauricio Faria de Oliveira-3
From: "Guilherme G. Piccoli" <[hidden email]>

BugLink: https://bugs.launchpad.net/bugs/1797990

Recently was noticed in an HP GEN9 system that kdump couldn't succeed
due to an irq storm coming from an Intel NIC, narrowed down to be lack
of clearing the MSI/MSI-X enable bits during the kdump kernel boot.
For that, we need an early quirk to manually turn off MSI/MSI-X for
PCI devices - this was worked as an optional boot parameter in a
(~subsequent~) previous patch.

Problem is that in our test system, the Intel NICs were not present in
any secondary bus under the first PCIe root complex, so they couldn't
be reached by the recursion in check_dev_quirk(). Modern systems,
specially with multi-processors and multiple NUMA nodes expose multiple
root complexes, describing more than one PCI hierarchy domain. Currently
the simple recursion present in the early-quirks code from x86 starts a
descending recursion from bus 0000:00, and reach many other busses by
navigating this hierarchy walking through the bridges. This is not
enough in systems with more than one root complex/host bridge, since
the recursion won't "traverse" to other root complexes by starting
statically in 0000:00 (for more details, see [0]).

This patch hence implements the full bus/device/function scan in
early_quirks(), by checking all possible busses instead of using a
recursion based on the first root bus or limiting the search scope to
the first 32 busses (like it was done in the beginning [1]).

[0] https://bugs.launchpad.net/bugs/1797990

[1] From historical perspective, early PCI scan dates back
to BitKeeper, added by Andi Kleen's "[PATCH] APIC fixes for x86-64",
on October/2003. It initially restricted the search to the first
32 busses and slots.

Due to a potential bug found in Nvidia chipsets, the scan
was changed to run only in the first root bus: see
commit 8659c406ade3 ("x86: only scan the root bus in early PCI quirks")

Finally, secondary busses reachable from the 1st bus were re-added back by:
commit 850c321027c2 ("x86/quirks: Reintroduce scanning of secondary buses")

Reported-by: Dan Streetman <[hidden email]>
Signed-off-by: Guilherme G. Piccoli <[hidden email]>
[mfo: v2:
 - gate the bus-scan differences with the cmdline option.
 - update changelog: subsequent/previous patch.]
Signed-off-by: Mauricio Faria de Oliveira <[hidden email]>
---
 arch/x86/kernel/early-quirks.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/arch/x86/kernel/early-quirks.c b/arch/x86/kernel/early-quirks.c
index e7561314eb5d..330749614db6 100644
--- a/arch/x86/kernel/early-quirks.c
+++ b/arch/x86/kernel/early-quirks.c
@@ -793,6 +793,10 @@ static int __init check_dev_quirk(int num, int slot, int func)
     PCI_HEADER_TYPE);
 
  if ((type & 0x7f) == PCI_HEADER_TYPE_BRIDGE) {
+ /* pci_early_clear_msi scans the buses differently. */
+ if (pci_early_clear_msi)
+ return -1;
+
  sec = read_pci_config_byte(num, slot, func, PCI_SECONDARY_BUS);
  if (sec > num)
  early_pci_scan_bus(sec);
@@ -819,8 +823,13 @@ static void __init early_pci_scan_bus(int bus)
 
 void __init early_quirks(void)
 {
+ int bus;
+
  if (!early_pci_allowed())
  return;
 
  early_pci_scan_bus(0);
+ /* pci_early_clear_msi scans more buses. */
+ for (bus = 1; pci_early_clear_msi && bus < 256; bus++)
+ early_pci_scan_bus(bus);
 }
--
2.17.1


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

ACK: [D][PATCH v2 0/3] Add kernel parameter 'pci=clearmsi' to clear MSI(X)s early on boot

Thadeu Lima de Souza Cascardo-3
In reply to this post by Mauricio Faria de Oliveira-3
Same considerations that I had for Xenial.

--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

Re: [D][PATCH v2 0/3] Add kernel parameter 'pci=clearmsi' to clear MSI(X)s early on boot

Thadeu Lima de Souza Cascardo-3
In reply to this post by Mauricio Faria de Oliveira-3
Hi, Mauricio.

So, disco master-next does not correspond yet to our unstable development,
which is already on top of 4.19. The patchset does not apply cleanly to
unstable master branch. Can you submit for unstable, then?

Thanks.
Cascardo.

--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

Re: [D][PATCH v2 0/3] Add kernel parameter 'pci=clearmsi' to clear MSI(X)s early on boot

Mauricio Faria de Oliveira-3
On Thu, Nov 8, 2018 at 3:21 PM Thadeu Lima de Souza Cascardo <[hidden email]> wrote:
Hi, Mauricio.

So, disco master-next does not correspond yet to our unstable development,
which is already on top of 4.19. The patchset does not apply cleanly to
unstable master branch. Can you submit for unstable, then?

Sure; it will take a little while to build/test, but it should hit the list today as well.


Thanks.
Cascardo.


--
Mauricio Faria de Oliveira

--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team