This simple patch set fixed some serious security issues found when DPC
error injection and NVMe SSD hotplug brute force test were doing -- race
condition between DPC handler and pciehp, AER interrupt handlers, caused
system hang and system with DPC feature couldn't recover to normal
working state as expected (NVMe instance lost, mount operation hang,
race PCIe access caused uncorrectable errors reported alternatively etc).
With this patch set applied, stable 5.9-rc6 on ICS (Ice Lake SP platform,
see
https://en.wikichip.org/wiki/intel/microarchitectures/ice_lake_(server))
could pass the PCIe Gen4 NVMe SSD brute force hotplug test with any time
interval between hot-remove and plug-in operation tens of times without
any errors occur and system works normal.
With this patch set applied, system with DPC feature could recover from
NON-FATAL and FATAL errors injection test and works as expected.
System works smoothly when errors happen while hotplug is doing, no
uncorrectable errors found.
Brute DPC error injection script:
for i in {0..100}
do
setpci -s 64:02.0 0x196.w=000a
setpci -s 65:00.0 0x04.w=0544
mount /dev/nvme0n1p1 /root/nvme
sleep 1
done
Other details see every commits description part.
This patch set could be applied to stable 5.9-rc6 directly.
Help to review and test.
V2: changed according to review by Andy Shevchenko.
Thanks,
Ethan
Ethan Zhao (5):
PCI: define a function to check and wait till port finish DPC handling
PCI: pciehp: check and wait port status out of DPC before handling
DLLSC and PDC
PCI/ERR: get device before call device driver to avoid NULL pointer
reference
PCI: only return true when dev io state is really changed
PCI/ERR: don't mix io state not changed and no driver together
drivers/pci/hotplug/pciehp_hpc.c | 4 +++-
drivers/pci/pci.h | 34 +++++---------------------------
drivers/pci/pcie/err.c | 18 +++++++++++++++--
include/linux/pci.h | 31 +++++++++++++++++++++++++++++
4 files changed, 55 insertions(+), 32 deletions(-)
--
2.18.4
When we see 'can't recover (no error_detected callback)' on console,
Maybe the reason is io state is not changed by calling
pci_dev_set_io_state(), that is confused. fix it.
Signed-off-by: Ethan Zhao <[email protected]>
Tested-by: Wen Jin <[email protected]>
Tested-by: Shanshan Zhang <[email protected]>
---
Chagnes:
V2: no change.
drivers/pci/pcie/err.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/drivers/pci/pcie/err.c b/drivers/pci/pcie/err.c
index e35c4480c86b..d85f27c90c26 100644
--- a/drivers/pci/pcie/err.c
+++ b/drivers/pci/pcie/err.c
@@ -55,8 +55,10 @@ static int report_error_detected(struct pci_dev *dev,
if (!pci_dev_get(dev))
return 0;
device_lock(&dev->dev);
- if (!pci_dev_set_io_state(dev, state) ||
- !dev->driver ||
+ if (!pci_dev_set_io_state(dev, state)) {
+ pci_dbg(dev, "Device might already being in error handling ...\n");
+ vote = PCI_ERS_RESULT_NONE;
+ } else if (!dev->driver ||
!dev->driver->err_handler ||
!dev->driver->err_handler->error_detected) {
/*
--
2.18.4
When uncorrectable error happens, AER driver and DPC driver interrupt
handlers likely call
pcie_do_recovery()
->pci_walk_bus()
->report_frozen_detected()
with pci_channel_io_frozen the same time.
If pci_dev_set_io_state() return true even if the original state is
pci_channel_io_frozen, that will cause AER or DPC handler re-enter
the error detecting and recovery procedure one after another.
The result is the recovery flow mixed between AER and DPC.
So simplify the pci_dev_set_io_state() function to only return true
when dev->error_state is changed.
Signed-off-by: Ethan Zhao <[email protected]>
Tested-by: Wen Jin <[email protected]>
Tested-by: Shanshan Zhang <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Reviewed-by: Alexandru Gagniuc <[email protected]>
---
Changes:
V2: revise doc and code flow according to Andy's suggestion.
drivers/pci/pci.h | 34 +++++-----------------------------
1 file changed, 5 insertions(+), 29 deletions(-)
diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
index fa12f7cbc1a0..387f891ce6a1 100644
--- a/drivers/pci/pci.h
+++ b/drivers/pci/pci.h
@@ -362,35 +362,11 @@ static inline bool pci_dev_set_io_state(struct pci_dev *dev,
bool changed = false;
device_lock_assert(&dev->dev);
- switch (new) {
- case pci_channel_io_perm_failure:
- switch (dev->error_state) {
- case pci_channel_io_frozen:
- case pci_channel_io_normal:
- case pci_channel_io_perm_failure:
- changed = true;
- break;
- }
- break;
- case pci_channel_io_frozen:
- switch (dev->error_state) {
- case pci_channel_io_frozen:
- case pci_channel_io_normal:
- changed = true;
- break;
- }
- break;
- case pci_channel_io_normal:
- switch (dev->error_state) {
- case pci_channel_io_frozen:
- case pci_channel_io_normal:
- changed = true;
- break;
- }
- break;
- }
- if (changed)
- dev->error_state = new;
+ if (dev->error_state == new)
+ return changed;
+
+ dev->error_state = new;
+ changed = true;
return changed;
}
--
2.18.4
During DPC error injection test we found there is race condition between
pciehp and DPC driver, NULL pointer reference caused panic as following
# setpci -s 64:02.0 0x196.w=000a
// 64:02.0 is rootport has DPC capability
# setpci -s 65:00.0 0x04.w=0544
// 65:00.0 is NVMe SSD populated in above port
# mount /dev/nvme0n1p1 nvme
(tested on stable 5.8 & ICS(Ice Lake SP platform, see
https://en.wikichip.org/wiki/intel/microarchitectures/ice_lake_(server))
Buffer I/O error on dev nvme0n1p1, logical block 468843328,
async page read
BUG: kernel NULL pointer dereference, address: 0000000000000050
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
PGD 0
Oops: 0000 [#1] SMP NOPTI
CPU: 12 PID: 513 Comm: irq/124-pcie-dp Not tainted 5.8.0-0.0.7.el8.x86_64+ #1
RIP: 0010:report_error_detected.cold.4+0x7d/0xe6
Code: b6 d0 e8 e8 fe 11 00 e8 16 c5 fb ff be 06 00 00 00 48 89 df e8 d3 65 ff
ff b8 06 00 00 00 e9 75 fc ff ff 48 8b 43 68 45 31 c9 <48> 8b 50 50 48 83 3a 00
41 0f 94 c1 45 31 c0 48 85 d2 41 0f 94 c0
RSP: 0018:ff8e06cf8762fda8 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ff4e3eaacf42a000 RCX: ff4e3eb31f223c01
RDX: ff4e3eaacf42a140 RSI: ff4e3eb31f223c00 RDI: ff4e3eaacf42a138
RBP: ff8e06cf8762fdd0 R08: 00000000000000bf R09: 0000000000000000
R10: 000000eb8ebeab53 R11: ffffffff93453258 R12: 0000000000000002
R13: ff4e3eaacf42a130 R14: ff8e06cf8762fe2c R15: ff4e3eab44733828
FS: 0000000000000000(0000) GS:ff4e3eab1fd00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000050 CR3: 0000000f8f80a004 CR4: 0000000000761ee0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
PKRU: 55555554
Call Trace:
? report_normal_detected+0x20/0x20
report_frozen_detected+0x16/0x20
pci_walk_bus+0x75/0x90
? dpc_irq+0x90/0x90
pcie_do_recovery+0x157/0x201
? irq_finalize_oneshot.part.47+0xe0/0xe0
dpc_handler+0x29/0x40
irq_thread_fn+0x24/0x60
irq_thread+0xea/0x170
? irq_forced_thread_fn+0x80/0x80
? irq_thread_check_affinity+0xf0/0xf0
kthread+0x124/0x140
? kthread_park+0x90/0x90
ret_from_fork+0x1f/0x30
Modules linked in: nft_fib_inet.........
CR2: 0000000000000050
Though we partly close the race condition with patch 'PCI: pciehp: check
and wait port status out of DPC before handling DLLSC and PDC', but there
is no hardware spec or software sequence to guarantee the pcie_ist() run
into pci_wait_port_outdpc() first or DPC triggered status bits being set
first when errors triggered DPC containment procedure, so device still
could be removed by function pci_stop_and_removed_bus_device() then freed
by pci_dev_put() in pciehp driver first during pcie_do_recover()/
pci_walk_bus() is called by dpc_handler() in DPC driver.
Maybe unify pci_bus_sem and pci_rescan_remove_lock to serialize the
removal and walking operation is the right way, but here we use
pci_dev_get() to increase the reference count of device before using the
device to avoid it is freed in use.
With this patch and patch 'PCI: pciehp: check and wait port status out of
DPC before handling DLLSC and PDC', stable 5.9-rc6 could pass the error
injection test and no panic happened.
Brute DPC error injection script:
for i in {0..100}
do
setpci -s 64:02.0 0x196.w=000a
setpci -s 65:00.0 0x04.w=0544
mount /dev/nvme0n1p1 /root/nvme
sleep 1
done
Signed-off-by: Ethan Zhao <[email protected]>
Tested-by: Wen Jin <[email protected]>
Tested-by: Shanshan Zhang <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
---
Changes:
V2: revise doc according to Andy's suggestion.
drivers/pci/pcie/err.c | 12 ++++++++++++
1 file changed, 12 insertions(+)
diff --git a/drivers/pci/pcie/err.c b/drivers/pci/pcie/err.c
index c543f419d8f9..e35c4480c86b 100644
--- a/drivers/pci/pcie/err.c
+++ b/drivers/pci/pcie/err.c
@@ -52,6 +52,8 @@ static int report_error_detected(struct pci_dev *dev,
pci_ers_result_t vote;
const struct pci_error_handlers *err_handler;
+ if (!pci_dev_get(dev))
+ return 0;
device_lock(&dev->dev);
if (!pci_dev_set_io_state(dev, state) ||
!dev->driver ||
@@ -76,6 +78,7 @@ static int report_error_detected(struct pci_dev *dev,
pci_uevent_ers(dev, vote);
*result = merge_result(*result, vote);
device_unlock(&dev->dev);
+ pci_dev_put(dev);
return 0;
}
@@ -94,6 +97,8 @@ static int report_mmio_enabled(struct pci_dev *dev, void *data)
pci_ers_result_t vote, *result = data;
const struct pci_error_handlers *err_handler;
+ if (!pci_dev_get(dev))
+ return 0;
device_lock(&dev->dev);
if (!dev->driver ||
!dev->driver->err_handler ||
@@ -105,6 +110,7 @@ static int report_mmio_enabled(struct pci_dev *dev, void *data)
*result = merge_result(*result, vote);
out:
device_unlock(&dev->dev);
+ pci_dev_put(dev);
return 0;
}
@@ -113,6 +119,8 @@ static int report_slot_reset(struct pci_dev *dev, void *data)
pci_ers_result_t vote, *result = data;
const struct pci_error_handlers *err_handler;
+ if (!pci_dev_get(dev))
+ return 0;
device_lock(&dev->dev);
if (!dev->driver ||
!dev->driver->err_handler ||
@@ -124,6 +132,7 @@ static int report_slot_reset(struct pci_dev *dev, void *data)
*result = merge_result(*result, vote);
out:
device_unlock(&dev->dev);
+ pci_dev_put(dev);
return 0;
}
@@ -131,6 +140,8 @@ static int report_resume(struct pci_dev *dev, void *data)
{
const struct pci_error_handlers *err_handler;
+ if (!pci_dev_get(dev))
+ return 0;
device_lock(&dev->dev);
if (!pci_dev_set_io_state(dev, pci_channel_io_normal) ||
!dev->driver ||
@@ -143,6 +154,7 @@ static int report_resume(struct pci_dev *dev, void *data)
out:
pci_uevent_ers(dev, PCI_ERS_RESULT_RECOVERED);
device_unlock(&dev->dev);
+ pci_dev_put(dev);
return 0;
}
--
2.18.4
On Sat, 2020-09-26 at 23:28 -0400, Ethan Zhao wrote:
> simplify the pci_dev_set_io_state() function to only return true
> when dev->error_state is changed.
[]
> diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
[]
> @@ -362,35 +362,11 @@ static inline bool pci_dev_set_io_state(struct pci_dev *dev,
> bool changed = false;
[]
> + if (dev->error_state == new)
> + return changed;
> +
> + dev->error_state = new;
> + changed = true;
> return changed;
> }
This would be simpler removing the unnecessary
changed automatic
...
if (dev->error_state == new)
return false;
dev->error_state = new;
return true;
}
When root port has DPC capability and it is enabled, then triggered by
errors, DPC DLLSC and PDC interrupts will be sent to DPC driver, pciehp
driver at the same time.
That will cause following result:
1. Link and device are recovered by hardware DPC and software DPC driver,
device
isn't removed, but the pciehp might treat it as device was hot removed.
2. Race condition happens bettween pciehp_unconfigure_device() called by
pciehp_ist() in pciehp driver and pci_do_recovery() called by
dpc_handler in DPC driver. no luck, there is no lock to protect
pci_stop_and_remove_bus_device()
against pci_walk_bus(), they hold different samphore and mutex,
pci_stop_and_remove_bus_device holds pci_rescan_remove_lock, and
pci_walk_bus() holds pci_bus_sem.
This race condition is not purely code analysis, it could be triggered by
following command series:
# setpci -s 64:02.0 0x196.w=000a // 64:02.0 rootport has DPC capability
# setpci -s 65:00.0 0x04.w=0544 // 65:00.0 NVMe SSD populated in port
# mount /dev/nvme0n1p1 nvme
One shot will cause system panic and NULL pointer reference happened.
(tested on stable 5.8 & ICS(Ice Lake SP platform, see
https://en.wikichip.org/wiki/intel/microarchitectures/ice_lake_(server))
Buffer I/O error on dev nvme0n1p1, logical block 3328, async page read
BUG: kernel NULL pointer dereference, address: 0000000000000050
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
PGD 0
Oops: 0000 [#1] SMP NOPTI
CPU: 12 PID: 513 Comm: irq/124-pcie-dp Not tainted 5.8.0 el8.x86_64+ #1
RIP: 0010:report_error_detected.cold.4+0x7d/0xe6
Code: b6 d0 e8 e8 fe 11 00 e8 16 c5 fb ff be 06 00 00 00 48 89 df e8 d3
65 ff ff b8 06 00 00 00 e9 75 fc ff ff 48 8b 43 68 45 31 c9 <48> 8b 50
50 48 83 3a 00 41 0f 94 c1 45 31 c0 48 85 d2 41 0f 94 c0
RSP: 0018:ff8e06cf8762fda8 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ff4e3eaacf42a000 RCX: ff4e3eb31f223c01
RDX: ff4e3eaacf42a140 RSI: ff4e3eb31f223c00 RDI: ff4e3eaacf42a138
RBP: ff8e06cf8762fdd0 R08: 00000000000000bf R09: 0000000000000000
R10: 000000eb8ebeab53 R11: ffffffff93453258 R12: 0000000000000002
R13: ff4e3eaacf42a130 R14: ff8e06cf8762fe2c R15: ff4e3eab44733828
FS: 0000000000000000(0000) GS:ff4e3eab1fd00000(0000) knl
GS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000050 CR3: 0000000f8f80a004 CR4: 0000000000761ee0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
PKRU: 55555554
Call Trace:
? report_normal_detected+0x20/0x20
report_frozen_detected+0x16/0x20
pci_walk_bus+0x75/0x90
? dpc_irq+0x90/0x90
pcie_do_recovery+0x157/0x201
? irq_finalize_oneshot.part.47+0xe0/0xe0
dpc_handler+0x29/0x40
irq_thread_fn+0x24/0x60
irq_thread+0xea/0x170
? irq_forced_thread_fn+0x80/0x80
? irq_thread_check_affinity+0xf0/0xf0
kthread+0x124/0x140
? kthread_park+0x90/0x90
ret_from_fork+0x1f/0x30
Modules linked in: nft_fib_inet.........
CR2: 0000000000000050
With this patch, the handling flow of DPC containment and hotplug is
partly ordered and serialized, let hardware DPC do the controller reset
etc recovery action first, then DPC driver handling the call-back from
device drivers, clear the DPC status, at the end, pciehp handle the DLLSC
and PDC etc.
Signed-off-by: Ethan Zhao <[email protected]>
Tested-by: Wen Jin <[email protected]>
Tested-by: Shanshan Zhang <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
---
Changes:
V2: revise doc according to Andy's suggestion.
drivers/pci/hotplug/pciehp_hpc.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/pci/hotplug/pciehp_hpc.c b/drivers/pci/hotplug/pciehp_hpc.c
index 53433b37e181..6f271160f18d 100644
--- a/drivers/pci/hotplug/pciehp_hpc.c
+++ b/drivers/pci/hotplug/pciehp_hpc.c
@@ -710,8 +710,10 @@ static irqreturn_t pciehp_ist(int irq, void *dev_id)
down_read(&ctrl->reset_lock);
if (events & DISABLE_SLOT)
pciehp_handle_disable_request(ctrl);
- else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC))
+ else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC)) {
+ pci_wait_port_outdpc(pdev);
pciehp_handle_presence_or_link_change(ctrl, events);
+ }
up_read(&ctrl->reset_lock);
ret = IRQ_HANDLED;
--
2.18.4
???? definitely simpler !
-----Original Message-----
From: Joe Perches <[email protected]>
Sent: Sunday, September 27, 2020 12:17 PM
To: Zhao, Haifeng <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
Cc: [email protected]; [email protected]; Jia, Pei P <[email protected]>; [email protected]; Kuppuswamy, Sathyanarayanan <[email protected]>
Subject: Re: [PATCH 4/5 V2] PCI: only return true when dev io state is really changed
On Sat, 2020-09-26 at 23:28 -0400, Ethan Zhao wrote:
> simplify the pci_dev_set_io_state() function to only return true when
> dev->error_state is changed.
[]
> diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
[]
> @@ -362,35 +362,11 @@ static inline bool pci_dev_set_io_state(struct pci_dev *dev,
> bool changed = false;
[]
> + if (dev->error_state == new)
> + return changed;
> +
> + dev->error_state = new;
> + changed = true;
> return changed;
> }
This would be simpler removing the unnecessary changed automatic
...
if (dev->error_state == new)
return false;
dev->error_state = new;
return true;
}
On 9/26/2020 11:28 PM, Ethan Zhao wrote:
> diff --git a/drivers/pci/hotplug/pciehp_hpc.c b/drivers/pci/hotplug/pciehp_hpc.c
> index 53433b37e181..6f271160f18d 100644
> --- a/drivers/pci/hotplug/pciehp_hpc.c
> +++ b/drivers/pci/hotplug/pciehp_hpc.c
> @@ -710,8 +710,10 @@ static irqreturn_t pciehp_ist(int irq, void *dev_id)
> down_read(&ctrl->reset_lock);
> if (events & DISABLE_SLOT)
> pciehp_handle_disable_request(ctrl);
> - else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC))
> + else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC)) {
> + pci_wait_port_outdpc(pdev);
> pciehp_handle_presence_or_link_change(ctrl, events);
> + }
> up_read(&ctrl->reset_lock);
This looks like a hack TBH.
Lukas, Keith;
What is your take on this?
Why is device lock not protecting this situation?
Is there a lock missing in hotplug driver?
Sinan
Sinan,
I explained the reason why locks don't protect this case in the patch description part.
Write side and read side hold different semaphore and mutex.
Thanks,
Ethan
-----Original Message-----
From: Sinan Kaya <[email protected]>
Sent: Sunday, September 27, 2020 11:28 PM
To: Zhao, Haifeng <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Keith Busch <[email protected]>
Cc: [email protected]; [email protected]; Jia, Pei P <[email protected]>; [email protected]; Kuppuswamy, Sathyanarayanan <[email protected]>
Subject: Re: [PATCH 2/5 V2] PCI: pciehp: check and wait port status out of DPC before handling DLLSC and PDC
On 9/26/2020 11:28 PM, Ethan Zhao wrote:
> diff --git a/drivers/pci/hotplug/pciehp_hpc.c b/drivers/pci/hotplug/pciehp_hpc.c
> index 53433b37e181..6f271160f18d 100644
> --- a/drivers/pci/hotplug/pciehp_hpc.c
> +++ b/drivers/pci/hotplug/pciehp_hpc.c
> @@ -710,8 +710,10 @@ static irqreturn_t pciehp_ist(int irq, void *dev_id)
> down_read(&ctrl->reset_lock);
> if (events & DISABLE_SLOT)
> pciehp_handle_disable_request(ctrl);
> - else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC))
> + else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC)) {
> + pci_wait_port_outdpc(pdev);
> pciehp_handle_presence_or_link_change(ctrl, events);
> + }
> up_read(&ctrl->reset_lock);
This looks like a hack TBH.
Lukas, Keith;
What is your take on this?
Why is device lock not protecting this situation?
Is there a lock missing in hotplug driver?
Sinan
On 9/27/2020 10:01 PM, Zhao, Haifeng wrote:
> Sinan,
> I explained the reason why locks don't protect this case in the patch description part.
> Write side and read side hold different semaphore and mutex.
>
I have been thinking about it some time but is there any reason why we
have to handle all port AER/DPC/HP events in different threads?
Can we go to single threaded event loop for all port drivers events?
This will require some refactoring but it wlll eliminate the lock
nightmares we are having.
This means no sleeping. All sleeps need to happen outside of the loop.
I wanted to see what you all are thinking about this.
It might become a performance problem if the system is
continuously observing a hotplug/aer/dpc events.
I always think that these should be rare events.
On 9/28/2020 7:10 AM, Sinan Kaya wrote:
> On 9/27/2020 10:01 PM, Zhao, Haifeng wrote:
>> Sinan,
>> I explained the reason why locks don't protect this case in the patch description part.
>> Write side and read side hold different semaphore and mutex.
>>
> I have been thinking about it some time but is there any reason why we
> have to handle all port AER/DPC/HP events in different threads?
>
> Can we go to single threaded event loop for all port drivers events?
>
> This will require some refactoring but it wlll eliminate the lock
> nightmares we are having.
>
> This means no sleeping. All sleeps need to happen outside of the loop.
>
> I wanted to see what you all are thinking about this.
>
> It might become a performance problem if the system is
> continuously observing a hotplug/aer/dpc events.
>
> I always think that these should be rare events.
If restructuring would be too costly, the preferred solution should be
to fix the locks in hotplug driver rather than throwing there a random
wait call.
On 9/28/20 9:43 AM, Sinan Kaya wrote:
> On 9/28/2020 7:10 AM, Sinan Kaya wrote:
>> On 9/27/2020 10:01 PM, Zhao, Haifeng wrote:
>>> Sinan,
>>> I explained the reason why locks don't protect this case in the patch description part.
>>> Write side and read side hold different semaphore and mutex.
>>>
>> I have been thinking about it some time but is there any reason why we
>> have to handle all port AER/DPC/HP events in different threads?
>>
>> Can we go to single threaded event loop for all port drivers events?
>>
>> This will require some refactoring but it wlll eliminate the lock
>> nightmares we are having.
>>
>> This means no sleeping. All sleeps need to happen outside of the loop.
>>
>> I wanted to see what you all are thinking about this.
>>
>> It might become a performance problem if the system is
>> continuously observing a hotplug/aer/dpc events.
>>
>> I always think that these should be rare events.
> If restructuring would be too costly, the preferred solution should be
> to fix the locks in hotplug driver rather than throwing there a random
> wait call.
Since the current race condition is detected between DPC and
hotplug, I recommend synchronizing them.
--
Sathyanarayanan Kuppuswamy
Linux Kernel Developer
On Tue, Sep 29, 2020 at 12:45 AM Kuppuswamy, Sathyanarayanan
<[email protected]> wrote:
>
>
> On 9/28/20 9:43 AM, Sinan Kaya wrote:
> > On 9/28/2020 7:10 AM, Sinan Kaya wrote:
> >> On 9/27/2020 10:01 PM, Zhao, Haifeng wrote:
> >>> Sinan,
> >>> I explained the reason why locks don't protect this case in the patch description part.
> >>> Write side and read side hold different semaphore and mutex.
> >>>
> >> I have been thinking about it some time but is there any reason why we
> >> have to handle all port AER/DPC/HP events in different threads?
> >>
> >> Can we go to single threaded event loop for all port drivers events?
> >>
> >> This will require some refactoring but it wlll eliminate the lock
> >> nightmares we are having.
> >>
> >> This means no sleeping. All sleeps need to happen outside of the loop.
> >>
> >> I wanted to see what you all are thinking about this.
> >>
> >> It might become a performance problem if the system is
> >> continuously observing a hotplug/aer/dpc events.
> >>
> >> I always think that these should be rare events.
> > If restructuring would be too costly, the preferred solution should be
> > to fix the locks in hotplug driver rather than throwing there a random
> > wait call.
> Since the current race condition is detected between DPC and
> hotplug, I recommend synchronizing them.
The locks are the first place to root cause and try to fix. but not so easy to
refactor the remove-scan-semaphore and the bus-walk-mutex. too expensive
work. --- rework every piece of code that uses them.
Thanks,
Ethan
>
> --
> Sathyanarayanan Kuppuswamy
> Linux Kernel Developer
>
On Tue, Sep 29, 2020 at 12:44 AM Sinan Kaya <[email protected]> wrote:
>
> On 9/28/2020 7:10 AM, Sinan Kaya wrote:
> > On 9/27/2020 10:01 PM, Zhao, Haifeng wrote:
> >> Sinan,
> >> I explained the reason why locks don't protect this case in the patch description part.
> >> Write side and read side hold different semaphore and mutex.
> >>
> > I have been thinking about it some time but is there any reason why we
> > have to handle all port AER/DPC/HP events in different threads?
> >
> > Can we go to single threaded event loop for all port drivers events?
> >
> > This will require some refactoring but it wlll eliminate the lock
> > nightmares we are having.
> >
> > This means no sleeping. All sleeps need to happen outside of the loop.
> >
> > I wanted to see what you all are thinking about this.
> >
> > It might become a performance problem if the system is
> > continuously observing a hotplug/aer/dpc events.
> >
> > I always think that these should be rare events.
>
> If restructuring would be too costly, the preferred solution should be
> to fix the locks in hotplug driver rather than throwing there a random
> wait call.
My first though is to unify the pci_bus_sem & pci_rescan_remove_lock
to one sleepable lock, but verifying every
locking scenario to sort out dead lock warning, it is horrible job. I
gave up and then played the device status waiting trick
to workaround it.
index 03d37128a24f..477d4c499f87 100644
--- a/drivers/pci/probe.c
+++ b/drivers/pci/probe.c
@@ -3223,17 +3223,19 @@ EXPORT_SYMBOL_GPL(pci_rescan_bus);
* pci_rescan_bus(), pci_rescan_bus_bridge_resize() and PCI device removal
* routines should always be executed under this mutex.
*/
-static DEFINE_MUTEX(pci_rescan_remove_lock);
+/* static DEFINE_MUTEX(pci_rescan_remove_lock); */
void pci_lock_rescan_remove(void)
{
- mutex_lock(&pci_rescan_remove_lock);
+ /*mutex_lock(&pci_rescan_remove_lock); */
+ down_write(&pci_bus_sem);
}
EXPORT_SYMBOL_GPL(pci_lock_rescan_remove);
void pci_unlock_rescan_remove(void)
{
- mutex_unlock(&pci_rescan_remove_lock);
+ /*mutex_unlock(&pci_rescan_remove_lock); */
+ up_write(&pci_bus_sem);
}
EXPORT_SYMBOL_GPL(pci_unlock_rescan_remove);
Thanks，
Ethan
On Sun, Sep 27, 2020 at 11:27:46AM -0400, Sinan Kaya wrote:
> On 9/26/2020 11:28 PM, Ethan Zhao wrote:
> > --- a/drivers/pci/hotplug/pciehp_hpc.c
> > +++ b/drivers/pci/hotplug/pciehp_hpc.c
> > @@ -710,8 +710,10 @@ static irqreturn_t pciehp_ist(int irq, void *dev_id)
> > down_read(&ctrl->reset_lock);
> > if (events & DISABLE_SLOT)
> > pciehp_handle_disable_request(ctrl);
> > - else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC))
> > + else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC)) {
> > + pci_wait_port_outdpc(pdev);
> > pciehp_handle_presence_or_link_change(ctrl, events);
> > + }
> > up_read(&ctrl->reset_lock);
>
> This looks like a hack TBH.
>
> Lukas, Keith;
>
> What is your take on this?
> Why is device lock not protecting this situation?
>
> Is there a lock missing in hotplug driver?
According to Ethan's commit message, there are two issues here:
One, that pciehp may remove a device even though DPC recovered the error,
and two, that a null pointer deref occurs.
The latter is most certainly not a locking issue but failure of DPC
to hold a reference on the pci_dev.
Thanks,
Lukas
On Tue, Sep 29, 2020 at 4:29 PM Lukas Wunner <[email protected]> wrote:
>
> On Sun, Sep 27, 2020 at 11:27:46AM -0400, Sinan Kaya wrote:
> > On 9/26/2020 11:28 PM, Ethan Zhao wrote:
> > > --- a/drivers/pci/hotplug/pciehp_hpc.c
> > > +++ b/drivers/pci/hotplug/pciehp_hpc.c
> > > @@ -710,8 +710,10 @@ static irqreturn_t pciehp_ist(int irq, void *dev_id)
> > > down_read(&ctrl->reset_lock);
> > > if (events & DISABLE_SLOT)
> > > pciehp_handle_disable_request(ctrl);
> > > - else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC))
> > > + else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC)) {
> > > + pci_wait_port_outdpc(pdev);
> > > pciehp_handle_presence_or_link_change(ctrl, events);
> > > + }
> > > up_read(&ctrl->reset_lock);
> >
> > This looks like a hack TBH.
> >
> > Lukas, Keith;
> >
> > What is your take on this?
> > Why is device lock not protecting this situation?
> >
> > Is there a lock missing in hotplug driver?
>
> According to Ethan's commit message, there are two issues here:
> One, that pciehp may remove a device even though DPC recovered the error,
> and two, that a null pointer deref occurs.
>
> The latter is most certainly not a locking issue but failure of DPC
> to hold a reference on the pci_dev.
This is what patch 3/5 proposed to fix. while this one is to re-order
the mixed DPC
recovery procedure and DLLSC/PDC event handling, to make pciehp to know the
exact recovered result of DPC to malfunctional device ---- link
recovered, still there,
or is removed from the slot.
Thanks,
Ethan
>
> Thanks,
>
> Lukas
On Tue, Sep 29, 2020 at 05:46:41PM +0800, Ethan Zhao wrote:
> On Tue, Sep 29, 2020 at 4:29 PM Lukas Wunner <[email protected]> wrote:
> > On Sun, Sep 27, 2020 at 11:27:46AM -0400, Sinan Kaya wrote:
> > > On 9/26/2020 11:28 PM, Ethan Zhao wrote:
> > > > --- a/drivers/pci/hotplug/pciehp_hpc.c
> > > > +++ b/drivers/pci/hotplug/pciehp_hpc.c
> > > > @@ -710,8 +710,10 @@ static irqreturn_t pciehp_ist(int irq, void *dev_id)
> > > > down_read(&ctrl->reset_lock);
> > > > if (events & DISABLE_SLOT)
> > > > pciehp_handle_disable_request(ctrl);
> > > > - else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC))
> > > > + else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC)) {
> > > > + pci_wait_port_outdpc(pdev);
> > > > pciehp_handle_presence_or_link_change(ctrl, events);
> > > > + }
> > > > up_read(&ctrl->reset_lock);
> > >
> > > This looks like a hack TBH.
[...]
> > > Why is device lock not protecting this situation?
> > > Is there a lock missing in hotplug driver?
> >
> > According to Ethan's commit message, there are two issues here:
> > One, that pciehp may remove a device even though DPC recovered the error,
> > and two, that a null pointer deref occurs.
> >
> > The latter is most certainly not a locking issue but failure of DPC
> > to hold a reference on the pci_dev.
>
> This is what patch 3/5 proposed to fix.
Please reorder the series to fix the null pointer deref first,
i.e. move patch 3 before patch 2. If the null pointer deref is
fixed by patch 3, do not mention it in patch 2.
Thanks,
Lukas
On Tue, Sep 29, 2020 at 6:08 PM Lukas Wunner <[email protected]> wrote:
>
> On Tue, Sep 29, 2020 at 05:46:41PM +0800, Ethan Zhao wrote:
> > On Tue, Sep 29, 2020 at 4:29 PM Lukas Wunner <[email protected]> wrote:
> > > On Sun, Sep 27, 2020 at 11:27:46AM -0400, Sinan Kaya wrote:
> > > > On 9/26/2020 11:28 PM, Ethan Zhao wrote:
> > > > > --- a/drivers/pci/hotplug/pciehp_hpc.c
> > > > > +++ b/drivers/pci/hotplug/pciehp_hpc.c
> > > > > @@ -710,8 +710,10 @@ static irqreturn_t pciehp_ist(int irq, void *dev_id)
> > > > > down_read(&ctrl->reset_lock);
> > > > > if (events & DISABLE_SLOT)
> > > > > pciehp_handle_disable_request(ctrl);
> > > > > - else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC))
> > > > > + else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC)) {
> > > > > + pci_wait_port_outdpc(pdev);
> > > > > pciehp_handle_presence_or_link_change(ctrl, events);
> > > > > + }
> > > > > up_read(&ctrl->reset_lock);
> > > >
> > > > This looks like a hack TBH.
> [...]
> > > > Why is device lock not protecting this situation?
> > > > Is there a lock missing in hotplug driver?
> > >
> > > According to Ethan's commit message, there are two issues here:
> > > One, that pciehp may remove a device even though DPC recovered the error,
> > > and two, that a null pointer deref occurs.
> > >
> > > The latter is most certainly not a locking issue but failure of DPC
> > > to hold a reference on the pci_dev.
> >
> > This is what patch 3/5 proposed to fix.
>
> Please reorder the series to fix the null pointer deref first,
> i.e. move patch 3 before patch 2. If the null pointer deref is
> fixed by patch 3, do not mention it in patch 2.
Make sense.
Thanks,
Ethan
>
> Thanks,
>
> Lukas