diff mbox series

[RFC,v3] PCI/portdrv: Only disable Bus Master on kexec reboot and connected PCI devices

Message ID 1600070215-3901-1-git-send-email-yangtiezhu@loongson.cn
State New
Delegated to: Bjorn Helgaas
Headers show
Series [RFC,v3] PCI/portdrv: Only disable Bus Master on kexec reboot and connected PCI devices | expand

Commit Message

Tiezhu Yang Sept. 14, 2020, 7:56 a.m. UTC
After commit 745be2e700cd ("PCIe: portdrv: call pci_disable_device
during remove") and commit cc27b735ad3a ("PCI/portdrv: Turn off PCIe
services during shutdown"), it also calls pci_disable_device() during
shutdown, this leads to shutdown or reboot failure occasionally due to
clear PCI_COMMAND_MASTER on the device in do_pci_disable_device().

drivers/pci/pci.c
static void do_pci_disable_device(struct pci_dev *dev)
{
        u16 pci_command;

        pci_read_config_word(dev, PCI_COMMAND, &pci_command);
        if (pci_command & PCI_COMMAND_MASTER) {
                pci_command &= ~PCI_COMMAND_MASTER;
                pci_write_config_word(dev, PCI_COMMAND, pci_command);
        }

        pcibios_disable_device(dev);
}

When remove "pci_command &= ~PCI_COMMAND_MASTER;", it can work well when
shutdown or reboot.

As Oliver O'Halloran said, no need to call pci_disable_device() when
actually shutting down, but we should call pci_disable_device() before
handing over to the new kernel on kexec reboot, so we can do some
condition checks which are already executed afterwards by the function
pci_device_shutdown(), this is done by commit 4fc9bbf98fd6 ("PCI: Disable
Bus Master only on kexec reboot") and commit 6e0eda3c3898 ("PCI: Don't try
to disable Bus Master on disconnected PCI devices").

drivers/pci/pci-driver.c
static void pci_device_shutdown(struct device *dev)
{
 ...
        if (drv && drv->shutdown)
                drv->shutdown(pci_dev);

        /*
         * If this is a kexec reboot, turn off Bus Master bit on the
         * device to tell it to not continue to do DMA. Don't touch
         * devices in D3cold or unknown states.
         * If it is not a kexec reboot, firmware will hit the PCI
         * devices with big hammer and stop their DMA any way.
         */
        if (kexec_in_progress && (pci_dev->current_state <= PCI_D3hot))
                pci_clear_master(pci_dev);
}

[   36.159446] Call Trace:
[   36.241688] [<ffffffff80211434>] show_stack+0x9c/0x130
[   36.326619] [<ffffffff80661b70>] dump_stack+0xb0/0xf0
[   36.410403] [<ffffffff806a8240>] pcie_portdrv_shutdown+0x18/0x78
[   36.495302] [<ffffffff8069c6b4>] pci_device_shutdown+0x44/0x90
[   36.580027] [<ffffffff807aac90>] device_shutdown+0x130/0x290
[   36.664486] [<ffffffff80265448>] kernel_power_off+0x38/0x80
[   36.748272] [<ffffffff80265634>] __do_sys_reboot+0x1a4/0x258
[   36.831985] [<ffffffff80218b90>] syscall_common+0x34/0x58

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
---
 drivers/pci/pcie/portdrv_core.c |  1 -
 drivers/pci/pcie/portdrv_pci.c  | 14 +++++++++++++-
 2 files changed, 13 insertions(+), 2 deletions(-)

Comments

Huacai Chen Sept. 14, 2020, 8:45 a.m. UTC | #1
Hi, Tiezhu,

How do you test kexec? kexec -e or systemctl kexec? Or both?

Huacai

陈华才江苏航天龙梦信息技术有限公司/研发中心/软件部   ------------------ Original ------------------From:  "Tiezhu Yang"<yangtiezhu@loongson.cn>;Date:  Mon, Sep 14, 2020 03:57 PMTo:  "Bjorn Helgaas"<bhelgaas@google.com>; Cc:  "linux-pci"<linux-pci@vger.kernel.org>; "linux-kernel"<linux-kernel@vger.kernel.org>; "Rafael J. Wysocki"<rafael.j.wysocki@intel.com>; "Konstantin Khlebnikov"<khlebnikov@openvz.org>; "Khalid Aziz"<khalid.aziz@oracle.com>; "Vivek Goyal"<vgoyal@redhat.com>; "Lukas Wunner"<lukas@wunner.de>; "Oliver O'Halloran"<oohall@gmail.com>; "Huacai Chen"<chenhc@lemote.com>; "Jiaxun Yang"<jiaxun.yang@flygoat.com>; "Xuefeng Li"<lixuefeng@loongson.cn>; Subject:  [RFC PATCH v3] PCI/portdrv: Only disable Bus Master on kexec reboot and connected PCI devices After commit 745be2e700cd ("PCIe: portdrv: call pci_disable_device
during remove") and commit cc27b735ad3a ("PCI/portdrv: Turn off PCIe
services during shutdown"), it also calls pci_disable_device() during
shutdown, this leads to shutdown or reboot failure occasionally due to
clear PCI_COMMAND_MASTER on the device in do_pci_disable_device().

drivers/pci/pci.c
static void do_pci_disable_device(struct pci_dev *dev)
{
        u16 pci_command;

        pci_read_config_word(dev, PCI_COMMAND, &pci_command);
        if (pci_command & PCI_COMMAND_MASTER) {
                pci_command &= ~PCI_COMMAND_MASTER;
                pci_write_config_word(dev, PCI_COMMAND, pci_command);
        }

        pcibios_disable_device(dev);
}

When remove "pci_command &= ~PCI_COMMAND_MASTER;", it can work well when
shutdown or reboot.

As Oliver O'Halloran said, no need to call pci_disable_device() when
actually shutting down, but we should call pci_disable_device() before
handing over to the new kernel on kexec reboot, so we can do some
condition checks which are already executed afterwards by the function
pci_device_shutdown(), this is done by commit 4fc9bbf98fd6 ("PCI: Disable
Bus Master only on kexec reboot") and commit 6e0eda3c3898 ("PCI: Don't try
to disable Bus Master on disconnected PCI devices").

drivers/pci/pci-driver.c
static void pci_device_shutdown(struct device *dev)
{
 ...
        if (drv && drv->shutdown)
                drv->shutdown(pci_dev);

        /*
         * If this is a kexec reboot, turn off Bus Master bit on the
         * device to tell it to not continue to do DMA. Don't touch
         * devices in D3cold or unknown states.
         * If it is not a kexec reboot, firmware will hit the PCI
         * devices with big hammer and stop their DMA any way.
         */
        if (kexec_in_progress && (pci_dev->current_state <= PCI_D3hot))
                pci_clear_master(pci_dev);
}

[   36.159446] Call Trace:
[   36.241688] [<ffffffff80211434>] show_stack+0x9c/0x130
[   36.326619] [<ffffffff80661b70>] dump_stack+0xb0/0xf0
[   36.410403] [<ffffffff806a8240>] pcie_portdrv_shutdown+0x18/0x78
[   36.495302] [<ffffffff8069c6b4>] pci_device_shutdown+0x44/0x90
[   36.580027] [<ffffffff807aac90>] device_shutdown+0x130/0x290
[   36.664486] [<ffffffff80265448>] kernel_power_off+0x38/0x80
[   36.748272] [<ffffffff80265634>] __do_sys_reboot+0x1a4/0x258
[   36.831985] [<ffffffff80218b90>] syscall_common+0x34/0x58

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
---
 drivers/pci/pcie/portdrv_core.c |  1 -
 drivers/pci/pcie/portdrv_pci.c  | 14 +++++++++++++-
 2 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/pcie/portdrv_core.c b/drivers/pci/pcie/portdrv_core.c
index 50a9522..1991aca 100644
--- a/drivers/pci/pcie/portdrv_core.c
+++ b/drivers/pci/pcie/portdrv_core.c
@@ -491,7 +491,6 @@ void pcie_port_device_remove(struct pci_dev *dev)
 {
 	device_for_each_child(&dev->dev, NULL, remove_iter);
 	pci_free_irq_vectors(dev);
-	pci_disable_device(dev);
 }
 
 /**
diff --git a/drivers/pci/pcie/portdrv_pci.c b/drivers/pci/pcie/portdrv_pci.c
index 3a3ce40..cab37a8 100644
--- a/drivers/pci/pcie/portdrv_pci.c
+++ b/drivers/pci/pcie/portdrv_pci.c
@@ -143,6 +143,18 @@ static void pcie_portdrv_remove(struct pci_dev *dev)
 	}
 
 	pcie_port_device_remove(dev);
+	pci_disable_device(dev);
+}
+
+static void pcie_portdrv_shutdown(struct pci_dev *dev)
+{
+	if (pci_bridge_d3_possible(dev)) {
+		pm_runtime_forbid(&dev->dev);
+		pm_runtime_get_noresume(&dev->dev);
+		pm_runtime_dont_use_autosuspend(&dev->dev);
+	}
+
+	pcie_port_device_remove(dev);
 }
 
 static pci_ers_result_t pcie_portdrv_error_detected(struct pci_dev *dev,
@@ -211,7 +223,7 @@ static struct pci_driver pcie_portdriver = {
 
 	.probe		= pcie_portdrv_probe,
 	.remove		= pcie_portdrv_remove,
-	.shutdown	= pcie_portdrv_remove,
+	.shutdown	= pcie_portdrv_shutdown,
 
 	.err_handler	= &pcie_portdrv_err_handler,
Huacai Chen Sept. 14, 2020, 8:52 a.m. UTC | #2
Hi, Tiezhu,

How do you test kexec? kexec -e or systemctl kexec? Or both?
P.S., Please also CC my gmail (chenhuacai@gmail.com) since lemote.com
has some communication problems.

Huacai

>
> 陈华才江苏航天龙梦信息技术有限公司/研发中心/软件部   ------------------ Original ------------------From:  "Tiezhu Yang"<yangtiezhu@loongson.cn>;Date:  Mon, Sep 14, 2020 03:57 PMTo:  "Bjorn Helgaas"<bhelgaas@google.com>; Cc:  "linux-pci"<linux-pci@vger.kernel.org>; "linux-kernel"<linux-kernel@vger.kernel.org>; "Rafael J. Wysocki"<rafael.j.wysocki@intel.com>; "Konstantin Khlebnikov"<khlebnikov@openvz.org>; "Khalid Aziz"<khalid.aziz@oracle.com>; "Vivek Goyal"<vgoyal@redhat.com>; "Lukas Wunner"<lukas@wunner.de>; "Oliver O'Halloran"<oohall@gmail.com>; "Huacai Chen"<chenhc@lemote.com>; "Jiaxun Yang"<jiaxun.yang@flygoat.com>; "Xuefeng Li"<lixuefeng@loongson.cn>; Subject:  [RFC PATCH v3] PCI/portdrv: Only disable Bus Master on kexec reboot and connected PCI devices After commit 745be2e700cd ("PCIe: portdrv: call pci_disable_device
> during remove") and commit cc27b735ad3a ("PCI/portdrv: Turn off PCIe
> services during shutdown"), it also calls pci_disable_device() during
> shutdown, this leads to shutdown or reboot failure occasionally due to
> clear PCI_COMMAND_MASTER on the device in do_pci_disable_device().
>
> drivers/pci/pci.c
> static void do_pci_disable_device(struct pci_dev *dev)
> {
>         u16 pci_command;
>
>         pci_read_config_word(dev, PCI_COMMAND, &pci_command);
>         if (pci_command & PCI_COMMAND_MASTER) {
>                 pci_command &= ~PCI_COMMAND_MASTER;
>                 pci_write_config_word(dev, PCI_COMMAND, pci_command);
>         }
>
>         pcibios_disable_device(dev);
> }
>
> When remove "pci_command &= ~PCI_COMMAND_MASTER;", it can work well when
> shutdown or reboot.
>
> As Oliver O'Halloran said, no need to call pci_disable_device() when
> actually shutting down, but we should call pci_disable_device() before
> handing over to the new kernel on kexec reboot, so we can do some
> condition checks which are already executed afterwards by the function
> pci_device_shutdown(), this is done by commit 4fc9bbf98fd6 ("PCI: Disable
> Bus Master only on kexec reboot") and commit 6e0eda3c3898 ("PCI: Don't try
> to disable Bus Master on disconnected PCI devices").
>
> drivers/pci/pci-driver.c
> static void pci_device_shutdown(struct device *dev)
> {
>  ...
>         if (drv && drv->shutdown)
>                 drv->shutdown(pci_dev);
>
>         /*
>          * If this is a kexec reboot, turn off Bus Master bit on the
>          * device to tell it to not continue to do DMA. Don't touch
>          * devices in D3cold or unknown states.
>          * If it is not a kexec reboot, firmware will hit the PCI
>          * devices with big hammer and stop their DMA any way.
>          */
>         if (kexec_in_progress && (pci_dev->current_state <= PCI_D3hot))
>                 pci_clear_master(pci_dev);
> }
>
> [   36.159446] Call Trace:
> [   36.241688] [<ffffffff80211434>] show_stack+0x9c/0x130
> [   36.326619] [<ffffffff80661b70>] dump_stack+0xb0/0xf0
> [   36.410403] [<ffffffff806a8240>] pcie_portdrv_shutdown+0x18/0x78
> [   36.495302] [<ffffffff8069c6b4>] pci_device_shutdown+0x44/0x90
> [   36.580027] [<ffffffff807aac90>] device_shutdown+0x130/0x290
> [   36.664486] [<ffffffff80265448>] kernel_power_off+0x38/0x80
> [   36.748272] [<ffffffff80265634>] __do_sys_reboot+0x1a4/0x258
> [   36.831985] [<ffffffff80218b90>] syscall_common+0x34/0x58
>
> Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
> ---
>  drivers/pci/pcie/portdrv_core.c |  1 -
>  drivers/pci/pcie/portdrv_pci.c  | 14 +++++++++++++-
>  2 files changed, 13 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/pci/pcie/portdrv_core.c b/drivers/pci/pcie/portdrv_core.c
> index 50a9522..1991aca 100644
> --- a/drivers/pci/pcie/portdrv_core.c
> +++ b/drivers/pci/pcie/portdrv_core.c
> @@ -491,7 +491,6 @@ void pcie_port_device_remove(struct pci_dev *dev)
>  {
>         device_for_each_child(&dev->dev, NULL, remove_iter);
>         pci_free_irq_vectors(dev);
> -       pci_disable_device(dev);
>  }
>
>  /**
> diff --git a/drivers/pci/pcie/portdrv_pci.c b/drivers/pci/pcie/portdrv_pci.c
> index 3a3ce40..cab37a8 100644
> --- a/drivers/pci/pcie/portdrv_pci.c
> +++ b/drivers/pci/pcie/portdrv_pci.c
> @@ -143,6 +143,18 @@ static void pcie_portdrv_remove(struct pci_dev *dev)
>         }
>
>         pcie_port_device_remove(dev);
> +       pci_disable_device(dev);
> +}
> +
> +static void pcie_portdrv_shutdown(struct pci_dev *dev)
> +{
> +       if (pci_bridge_d3_possible(dev)) {
> +               pm_runtime_forbid(&dev->dev);
> +               pm_runtime_get_noresume(&dev->dev);
> +               pm_runtime_dont_use_autosuspend(&dev->dev);
> +       }
> +
> +       pcie_port_device_remove(dev);
>  }
>
>  static pci_ers_result_t pcie_portdrv_error_detected(struct pci_dev *dev,
> @@ -211,7 +223,7 @@ static struct pci_driver pcie_portdriver = {
>
>         .probe          = pcie_portdrv_probe,
>         .remove         = pcie_portdrv_remove,
> -       .shutdown       = pcie_portdrv_remove,
> +       .shutdown       = pcie_portdrv_shutdown,
>
>         .err_handler    = &pcie_portdrv_err_handler,
>
> --
> 2.1.0
Tiezhu Yang Sept. 14, 2020, 9:29 a.m. UTC | #3
On 09/14/2020 04:52 PM, Huacai Chen wrote:
> Hi, Tiezhu,
>
> How do you test kexec? kexec -e or systemctl kexec? Or both?

kexec -l vmlinux --append="root=/dev/sda2 console=ttyS0,115200"
kexec -e

> P.S., Please also CC my gmail (chenhuacai@gmail.com) since lemote.com
> has some communication problems.

OK, no problem.

>
> Huacai
>
>> 陈华才江苏航天龙梦信息技术有限公司/研发中心/软件部   ------------------ Original ------------------From:  "Tiezhu Yang"<yangtiezhu@loongson.cn>;Date:  Mon, Sep 14, 2020 03:57 PMTo:  "Bjorn Helgaas"<bhelgaas@google.com>; Cc:  "linux-pci"<linux-pci@vger.kernel.org>; "linux-kernel"<linux-kernel@vger.kernel.org>; "Rafael J. Wysocki"<rafael.j.wysocki@intel.com>; "Konstantin Khlebnikov"<khlebnikov@openvz.org>; "Khalid Aziz"<khalid.aziz@oracle.com>; "Vivek Goyal"<vgoyal@redhat.com>; "Lukas Wunner"<lukas@wunner.de>; "Oliver O'Halloran"<oohall@gmail.com>; "Huacai Chen"<chenhc@lemote.com>; "Jiaxun Yang"<jiaxun.yang@flygoat.com>; "Xuefeng Li"<lixuefeng@loongson.cn>; Subject:  [RFC PATCH v3] PCI/portdrv: Only disable Bus Master on kexec reboot and connected PCI devices After commit 745be2e700cd ("PCIe: portdrv: call pci_disable_device
>> during remove") and commit cc27b735ad3a ("PCI/portdrv: Turn off PCIe
>> services during shutdown"), it also calls pci_disable_device() during
>> shutdown, this leads to shutdown or reboot failure occasionally due to
>> clear PCI_COMMAND_MASTER on the device in do_pci_disable_device().
>>
>> drivers/pci/pci.c
>> static void do_pci_disable_device(struct pci_dev *dev)
>> {
>>          u16 pci_command;
>>
>>          pci_read_config_word(dev, PCI_COMMAND, &pci_command);
>>          if (pci_command & PCI_COMMAND_MASTER) {
>>                  pci_command &= ~PCI_COMMAND_MASTER;
>>                  pci_write_config_word(dev, PCI_COMMAND, pci_command);
>>          }
>>
>>          pcibios_disable_device(dev);
>> }
>>
>> When remove "pci_command &= ~PCI_COMMAND_MASTER;", it can work well when
>> shutdown or reboot.
>>
>> As Oliver O'Halloran said, no need to call pci_disable_device() when
>> actually shutting down, but we should call pci_disable_device() before
>> handing over to the new kernel on kexec reboot, so we can do some
>> condition checks which are already executed afterwards by the function
>> pci_device_shutdown(), this is done by commit 4fc9bbf98fd6 ("PCI: Disable
>> Bus Master only on kexec reboot") and commit 6e0eda3c3898 ("PCI: Don't try
>> to disable Bus Master on disconnected PCI devices").
>>
>> drivers/pci/pci-driver.c
>> static void pci_device_shutdown(struct device *dev)
>> {
>>   ...
>>          if (drv && drv->shutdown)
>>                  drv->shutdown(pci_dev);
>>
>>          /*
>>           * If this is a kexec reboot, turn off Bus Master bit on the
>>           * device to tell it to not continue to do DMA. Don't touch
>>           * devices in D3cold or unknown states.
>>           * If it is not a kexec reboot, firmware will hit the PCI
>>           * devices with big hammer and stop their DMA any way.
>>           */
>>          if (kexec_in_progress && (pci_dev->current_state <= PCI_D3hot))
>>                  pci_clear_master(pci_dev);
>> }
>>
>> [   36.159446] Call Trace:
>> [   36.241688] [<ffffffff80211434>] show_stack+0x9c/0x130
>> [   36.326619] [<ffffffff80661b70>] dump_stack+0xb0/0xf0
>> [   36.410403] [<ffffffff806a8240>] pcie_portdrv_shutdown+0x18/0x78
>> [   36.495302] [<ffffffff8069c6b4>] pci_device_shutdown+0x44/0x90
>> [   36.580027] [<ffffffff807aac90>] device_shutdown+0x130/0x290
>> [   36.664486] [<ffffffff80265448>] kernel_power_off+0x38/0x80
>> [   36.748272] [<ffffffff80265634>] __do_sys_reboot+0x1a4/0x258
>> [   36.831985] [<ffffffff80218b90>] syscall_common+0x34/0x58
>>
>> Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
>> ---
>>   drivers/pci/pcie/portdrv_core.c |  1 -
>>   drivers/pci/pcie/portdrv_pci.c  | 14 +++++++++++++-
>>   2 files changed, 13 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/pci/pcie/portdrv_core.c b/drivers/pci/pcie/portdrv_core.c
>> index 50a9522..1991aca 100644
>> --- a/drivers/pci/pcie/portdrv_core.c
>> +++ b/drivers/pci/pcie/portdrv_core.c
>> @@ -491,7 +491,6 @@ void pcie_port_device_remove(struct pci_dev *dev)
>>   {
>>          device_for_each_child(&dev->dev, NULL, remove_iter);
>>          pci_free_irq_vectors(dev);
>> -       pci_disable_device(dev);
>>   }
>>
>>   /**
>> diff --git a/drivers/pci/pcie/portdrv_pci.c b/drivers/pci/pcie/portdrv_pci.c
>> index 3a3ce40..cab37a8 100644
>> --- a/drivers/pci/pcie/portdrv_pci.c
>> +++ b/drivers/pci/pcie/portdrv_pci.c
>> @@ -143,6 +143,18 @@ static void pcie_portdrv_remove(struct pci_dev *dev)
>>          }
>>
>>          pcie_port_device_remove(dev);
>> +       pci_disable_device(dev);
>> +}
>> +
>> +static void pcie_portdrv_shutdown(struct pci_dev *dev)
>> +{
>> +       if (pci_bridge_d3_possible(dev)) {
>> +               pm_runtime_forbid(&dev->dev);
>> +               pm_runtime_get_noresume(&dev->dev);
>> +               pm_runtime_dont_use_autosuspend(&dev->dev);
>> +       }
>> +
>> +       pcie_port_device_remove(dev);
>>   }
>>
>>   static pci_ers_result_t pcie_portdrv_error_detected(struct pci_dev *dev,
>> @@ -211,7 +223,7 @@ static struct pci_driver pcie_portdriver = {
>>
>>          .probe          = pcie_portdrv_probe,
>>          .remove         = pcie_portdrv_remove,
>> -       .shutdown       = pcie_portdrv_remove,
>> +       .shutdown       = pcie_portdrv_shutdown,
>>
>>          .err_handler    = &pcie_portdrv_err_handler,
>>
>> --
>> 2.1.0
Huacai Chen Sept. 14, 2020, 9:46 a.m. UTC | #4
Hi, Tiezhu,

On Mon, Sep 14, 2020 at 5:30 PM Tiezhu Yang <yangtiezhu@loongson.cn> wrote:
>
> On 09/14/2020 04:52 PM, Huacai Chen wrote:
> > Hi, Tiezhu,
> >
> > How do you test kexec? kexec -e or systemctl kexec? Or both?
>
> kexec -l vmlinux --append="root=/dev/sda2 console=ttyS0,115200"
> kexec -e
So you haven't tested "systemctl kexec"?

Huacai
>
> > P.S., Please also CC my gmail (chenhuacai@gmail.com) since lemote.com
> > has some communication problems.
>
> OK, no problem.
>
> >
> > Huacai
> >
> >> 陈华才江苏航天龙梦信息技术有限公司/研发中心/软件部   ------------------ Original ------------------From:  "Tiezhu Yang"<yangtiezhu@loongson.cn>;Date:  Mon, Sep 14, 2020 03:57 PMTo:  "Bjorn Helgaas"<bhelgaas@google.com>; Cc:  "linux-pci"<linux-pci@vger.kernel.org>; "linux-kernel"<linux-kernel@vger.kernel.org>; "Rafael J. Wysocki"<rafael.j.wysocki@intel.com>; "Konstantin Khlebnikov"<khlebnikov@openvz.org>; "Khalid Aziz"<khalid.aziz@oracle.com>; "Vivek Goyal"<vgoyal@redhat.com>; "Lukas Wunner"<lukas@wunner.de>; "Oliver O'Halloran"<oohall@gmail.com>; "Huacai Chen"<chenhc@lemote.com>; "Jiaxun Yang"<jiaxun.yang@flygoat.com>; "Xuefeng Li"<lixuefeng@loongson.cn>; Subject:  [RFC PATCH v3] PCI/portdrv: Only disable Bus Master on kexec reboot and connected PCI devices After commit 745be2e700cd ("PCIe: portdrv: call pci_disable_device
> >> during remove") and commit cc27b735ad3a ("PCI/portdrv: Turn off PCIe
> >> services during shutdown"), it also calls pci_disable_device() during
> >> shutdown, this leads to shutdown or reboot failure occasionally due to
> >> clear PCI_COMMAND_MASTER on the device in do_pci_disable_device().
> >>
> >> drivers/pci/pci.c
> >> static void do_pci_disable_device(struct pci_dev *dev)
> >> {
> >>          u16 pci_command;
> >>
> >>          pci_read_config_word(dev, PCI_COMMAND, &pci_command);
> >>          if (pci_command & PCI_COMMAND_MASTER) {
> >>                  pci_command &= ~PCI_COMMAND_MASTER;
> >>                  pci_write_config_word(dev, PCI_COMMAND, pci_command);
> >>          }
> >>
> >>          pcibios_disable_device(dev);
> >> }
> >>
> >> When remove "pci_command &= ~PCI_COMMAND_MASTER;", it can work well when
> >> shutdown or reboot.
> >>
> >> As Oliver O'Halloran said, no need to call pci_disable_device() when
> >> actually shutting down, but we should call pci_disable_device() before
> >> handing over to the new kernel on kexec reboot, so we can do some
> >> condition checks which are already executed afterwards by the function
> >> pci_device_shutdown(), this is done by commit 4fc9bbf98fd6 ("PCI: Disable
> >> Bus Master only on kexec reboot") and commit 6e0eda3c3898 ("PCI: Don't try
> >> to disable Bus Master on disconnected PCI devices").
> >>
> >> drivers/pci/pci-driver.c
> >> static void pci_device_shutdown(struct device *dev)
> >> {
> >>   ...
> >>          if (drv && drv->shutdown)
> >>                  drv->shutdown(pci_dev);
> >>
> >>          /*
> >>           * If this is a kexec reboot, turn off Bus Master bit on the
> >>           * device to tell it to not continue to do DMA. Don't touch
> >>           * devices in D3cold or unknown states.
> >>           * If it is not a kexec reboot, firmware will hit the PCI
> >>           * devices with big hammer and stop their DMA any way.
> >>           */
> >>          if (kexec_in_progress && (pci_dev->current_state <= PCI_D3hot))
> >>                  pci_clear_master(pci_dev);
> >> }
> >>
> >> [   36.159446] Call Trace:
> >> [   36.241688] [<ffffffff80211434>] show_stack+0x9c/0x130
> >> [   36.326619] [<ffffffff80661b70>] dump_stack+0xb0/0xf0
> >> [   36.410403] [<ffffffff806a8240>] pcie_portdrv_shutdown+0x18/0x78
> >> [   36.495302] [<ffffffff8069c6b4>] pci_device_shutdown+0x44/0x90
> >> [   36.580027] [<ffffffff807aac90>] device_shutdown+0x130/0x290
> >> [   36.664486] [<ffffffff80265448>] kernel_power_off+0x38/0x80
> >> [   36.748272] [<ffffffff80265634>] __do_sys_reboot+0x1a4/0x258
> >> [   36.831985] [<ffffffff80218b90>] syscall_common+0x34/0x58
> >>
> >> Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
> >> ---
> >>   drivers/pci/pcie/portdrv_core.c |  1 -
> >>   drivers/pci/pcie/portdrv_pci.c  | 14 +++++++++++++-
> >>   2 files changed, 13 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/drivers/pci/pcie/portdrv_core.c b/drivers/pci/pcie/portdrv_core.c
> >> index 50a9522..1991aca 100644
> >> --- a/drivers/pci/pcie/portdrv_core.c
> >> +++ b/drivers/pci/pcie/portdrv_core.c
> >> @@ -491,7 +491,6 @@ void pcie_port_device_remove(struct pci_dev *dev)
> >>   {
> >>          device_for_each_child(&dev->dev, NULL, remove_iter);
> >>          pci_free_irq_vectors(dev);
> >> -       pci_disable_device(dev);
> >>   }
> >>
> >>   /**
> >> diff --git a/drivers/pci/pcie/portdrv_pci.c b/drivers/pci/pcie/portdrv_pci.c
> >> index 3a3ce40..cab37a8 100644
> >> --- a/drivers/pci/pcie/portdrv_pci.c
> >> +++ b/drivers/pci/pcie/portdrv_pci.c
> >> @@ -143,6 +143,18 @@ static void pcie_portdrv_remove(struct pci_dev *dev)
> >>          }
> >>
> >>          pcie_port_device_remove(dev);
> >> +       pci_disable_device(dev);
> >> +}
> >> +
> >> +static void pcie_portdrv_shutdown(struct pci_dev *dev)
> >> +{
> >> +       if (pci_bridge_d3_possible(dev)) {
> >> +               pm_runtime_forbid(&dev->dev);
> >> +               pm_runtime_get_noresume(&dev->dev);
> >> +               pm_runtime_dont_use_autosuspend(&dev->dev);
> >> +       }
> >> +
> >> +       pcie_port_device_remove(dev);
> >>   }
> >>
> >>   static pci_ers_result_t pcie_portdrv_error_detected(struct pci_dev *dev,
> >> @@ -211,7 +223,7 @@ static struct pci_driver pcie_portdriver = {
> >>
> >>          .probe          = pcie_portdrv_probe,
> >>          .remove         = pcie_portdrv_remove,
> >> -       .shutdown       = pcie_portdrv_remove,
> >> +       .shutdown       = pcie_portdrv_shutdown,
> >>
> >>          .err_handler    = &pcie_portdrv_err_handler,
> >>
> >> --
> >> 2.1.0
>
Tiezhu Yang Sept. 14, 2020, 11:24 a.m. UTC | #5
On 09/14/2020 05:46 PM, Huacai Chen wrote:
> Hi, Tiezhu,
>
> On Mon, Sep 14, 2020 at 5:30 PM Tiezhu Yang <yangtiezhu@loongson.cn> wrote:
>> On 09/14/2020 04:52 PM, Huacai Chen wrote:
>>> Hi, Tiezhu,
>>>
>>> How do you test kexec? kexec -e or systemctl kexec? Or both?
>> kexec -l vmlinux --append="root=/dev/sda2 console=ttyS0,115200"
>> kexec -e
> So you haven't tested "systemctl kexec"?

Yes, the distro I used is Loongnix which has not kexec service now.
Is there any problem when use systemctl kexec? If you have more details,
please let me know.

>
> Huacai
>>> P.S., Please also CC my gmail (chenhuacai@gmail.com) since lemote.com
>>> has some communication problems.
>> OK, no problem.
>>
>>> Huacai
>>>
>>>> 陈华才江苏航天龙梦信息技术有限公司/研发中心/软件部   ------------------ Original ------------------From:  "Tiezhu Yang"<yangtiezhu@loongson.cn>;Date:  Mon, Sep 14, 2020 03:57 PMTo:  "Bjorn Helgaas"<bhelgaas@google.com>; Cc:  "linux-pci"<linux-pci@vger.kernel.org>; "linux-kernel"<linux-kernel@vger.kernel.org>; "Rafael J. Wysocki"<rafael.j.wysocki@intel.com>; "Konstantin Khlebnikov"<khlebnikov@openvz.org>; "Khalid Aziz"<khalid.aziz@oracle.com>; "Vivek Goyal"<vgoyal@redhat.com>; "Lukas Wunner"<lukas@wunner.de>; "Oliver O'Halloran"<oohall@gmail.com>; "Huacai Chen"<chenhc@lemote.com>; "Jiaxun Yang"<jiaxun.yang@flygoat.com>; "Xuefeng Li"<lixuefeng@loongson.cn>; Subject:  [RFC PATCH v3] PCI/portdrv: Only disable Bus Master on kexec reboot and connected PCI devices After commit 745be2e700cd ("PCIe: portdrv: call pci_disable_device
>>>> during remove") and commit cc27b735ad3a ("PCI/portdrv: Turn off PCIe
>>>> services during shutdown"), it also calls pci_disable_device() during
>>>> shutdown, this leads to shutdown or reboot failure occasionally due to
>>>> clear PCI_COMMAND_MASTER on the device in do_pci_disable_device().
>>>>
>>>> drivers/pci/pci.c
>>>> static void do_pci_disable_device(struct pci_dev *dev)
>>>> {
>>>>           u16 pci_command;
>>>>
>>>>           pci_read_config_word(dev, PCI_COMMAND, &pci_command);
>>>>           if (pci_command & PCI_COMMAND_MASTER) {
>>>>                   pci_command &= ~PCI_COMMAND_MASTER;
>>>>                   pci_write_config_word(dev, PCI_COMMAND, pci_command);
>>>>           }
>>>>
>>>>           pcibios_disable_device(dev);
>>>> }
>>>>
>>>> When remove "pci_command &= ~PCI_COMMAND_MASTER;", it can work well when
>>>> shutdown or reboot.
>>>>
>>>> As Oliver O'Halloran said, no need to call pci_disable_device() when
>>>> actually shutting down, but we should call pci_disable_device() before
>>>> handing over to the new kernel on kexec reboot, so we can do some
>>>> condition checks which are already executed afterwards by the function
>>>> pci_device_shutdown(), this is done by commit 4fc9bbf98fd6 ("PCI: Disable
>>>> Bus Master only on kexec reboot") and commit 6e0eda3c3898 ("PCI: Don't try
>>>> to disable Bus Master on disconnected PCI devices").
>>>>
>>>> drivers/pci/pci-driver.c
>>>> static void pci_device_shutdown(struct device *dev)
>>>> {
>>>>    ...
>>>>           if (drv && drv->shutdown)
>>>>                   drv->shutdown(pci_dev);
>>>>
>>>>           /*
>>>>            * If this is a kexec reboot, turn off Bus Master bit on the
>>>>            * device to tell it to not continue to do DMA. Don't touch
>>>>            * devices in D3cold or unknown states.
>>>>            * If it is not a kexec reboot, firmware will hit the PCI
>>>>            * devices with big hammer and stop their DMA any way.
>>>>            */
>>>>           if (kexec_in_progress && (pci_dev->current_state <= PCI_D3hot))
>>>>                   pci_clear_master(pci_dev);
>>>> }
>>>>
>>>> [   36.159446] Call Trace:
>>>> [   36.241688] [<ffffffff80211434>] show_stack+0x9c/0x130
>>>> [   36.326619] [<ffffffff80661b70>] dump_stack+0xb0/0xf0
>>>> [   36.410403] [<ffffffff806a8240>] pcie_portdrv_shutdown+0x18/0x78
>>>> [   36.495302] [<ffffffff8069c6b4>] pci_device_shutdown+0x44/0x90
>>>> [   36.580027] [<ffffffff807aac90>] device_shutdown+0x130/0x290
>>>> [   36.664486] [<ffffffff80265448>] kernel_power_off+0x38/0x80
>>>> [   36.748272] [<ffffffff80265634>] __do_sys_reboot+0x1a4/0x258
>>>> [   36.831985] [<ffffffff80218b90>] syscall_common+0x34/0x58
>>>>
>>>> Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
>>>> ---
>>>>    drivers/pci/pcie/portdrv_core.c |  1 -
>>>>    drivers/pci/pcie/portdrv_pci.c  | 14 +++++++++++++-
>>>>    2 files changed, 13 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/drivers/pci/pcie/portdrv_core.c b/drivers/pci/pcie/portdrv_core.c
>>>> index 50a9522..1991aca 100644
>>>> --- a/drivers/pci/pcie/portdrv_core.c
>>>> +++ b/drivers/pci/pcie/portdrv_core.c
>>>> @@ -491,7 +491,6 @@ void pcie_port_device_remove(struct pci_dev *dev)
>>>>    {
>>>>           device_for_each_child(&dev->dev, NULL, remove_iter);
>>>>           pci_free_irq_vectors(dev);
>>>> -       pci_disable_device(dev);
>>>>    }
>>>>
>>>>    /**
>>>> diff --git a/drivers/pci/pcie/portdrv_pci.c b/drivers/pci/pcie/portdrv_pci.c
>>>> index 3a3ce40..cab37a8 100644
>>>> --- a/drivers/pci/pcie/portdrv_pci.c
>>>> +++ b/drivers/pci/pcie/portdrv_pci.c
>>>> @@ -143,6 +143,18 @@ static void pcie_portdrv_remove(struct pci_dev *dev)
>>>>           }
>>>>
>>>>           pcie_port_device_remove(dev);
>>>> +       pci_disable_device(dev);
>>>> +}
>>>> +
>>>> +static void pcie_portdrv_shutdown(struct pci_dev *dev)
>>>> +{
>>>> +       if (pci_bridge_d3_possible(dev)) {
>>>> +               pm_runtime_forbid(&dev->dev);
>>>> +               pm_runtime_get_noresume(&dev->dev);
>>>> +               pm_runtime_dont_use_autosuspend(&dev->dev);
>>>> +       }
>>>> +
>>>> +       pcie_port_device_remove(dev);
>>>>    }
>>>>
>>>>    static pci_ers_result_t pcie_portdrv_error_detected(struct pci_dev *dev,
>>>> @@ -211,7 +223,7 @@ static struct pci_driver pcie_portdriver = {
>>>>
>>>>           .probe          = pcie_portdrv_probe,
>>>>           .remove         = pcie_portdrv_remove,
>>>> -       .shutdown       = pcie_portdrv_remove,
>>>> +       .shutdown       = pcie_portdrv_shutdown,
>>>>
>>>>           .err_handler    = &pcie_portdrv_err_handler,
>>>>
>>>> --
>>>> 2.1.0
周琰杰 (Zhou Yanjie) Sept. 14, 2020, 12:06 p.m. UTC | #6
Hello Tiezhu

在 2020/9/14 下午7:24, Tiezhu Yang 写道:
> On 09/14/2020 05:46 PM, Huacai Chen wrote:
>> Hi, Tiezhu,
>>
>> On Mon, Sep 14, 2020 at 5:30 PM Tiezhu Yang <yangtiezhu@loongson.cn> 
>> wrote:
>>> On 09/14/2020 04:52 PM, Huacai Chen wrote:
>>>> Hi, Tiezhu,
>>>>
>>>> How do you test kexec? kexec -e or systemctl kexec? Or both?
>>> kexec -l vmlinux --append="root=/dev/sda2 console=ttyS0,115200"
>>> kexec -e
>> So you haven't tested "systemctl kexec"?
>
> Yes, the distro I used is Loongnix which has not kexec service now.
> Is there any problem when use systemctl kexec? If you have more details,
> please let me know.
>

It is inevitable that there will be some omissions during testing. This 
is understandable, but what I am more curious about is why you send the 
patch that Huacai has already sent every time. Why not strengthen 
communication with Huacai and avoid duplication of work?


>>
>> Huacai
>>>> P.S., Please also CC my gmail (chenhuacai@gmail.com) since lemote.com
>>>> has some communication problems.
>>> OK, no problem.
>>>
>>>> Huacai
>>>>
>>>>> 陈华才江苏航天龙梦信息技术有限公司/研发中心/软件部 ------------------ 
>>>>> Original ------------------From: "Tiezhu 
>>>>> Yang"<yangtiezhu@loongson.cn>;Date:  Mon, Sep 14, 2020 03:57 
>>>>> PMTo:  "Bjorn Helgaas"<bhelgaas@google.com>; Cc: 
>>>>> "linux-pci"<linux-pci@vger.kernel.org>; 
>>>>> "linux-kernel"<linux-kernel@vger.kernel.org>; "Rafael J. 
>>>>> Wysocki"<rafael.j.wysocki@intel.com>; "Konstantin 
>>>>> Khlebnikov"<khlebnikov@openvz.org>; "Khalid 
>>>>> Aziz"<khalid.aziz@oracle.com>; "Vivek Goyal"<vgoyal@redhat.com>; 
>>>>> "Lukas Wunner"<lukas@wunner.de>; "Oliver 
>>>>> O'Halloran"<oohall@gmail.com>; "Huacai Chen"<chenhc@lemote.com>; 
>>>>> "Jiaxun Yang"<jiaxun.yang@flygoat.com>; "Xuefeng 
>>>>> Li"<lixuefeng@loongson.cn>; Subject:  [RFC PATCH v3] PCI/portdrv: 
>>>>> Only disable Bus Master on kexec reboot and connected PCI devices 
>>>>> After commit 745be2e700cd ("PCIe: portdrv: call pci_disable_device
>>>>> during remove") and commit cc27b735ad3a ("PCI/portdrv: Turn off PCIe
>>>>> services during shutdown"), it also calls pci_disable_device() during
>>>>> shutdown, this leads to shutdown or reboot failure occasionally 
>>>>> due to
>>>>> clear PCI_COMMAND_MASTER on the device in do_pci_disable_device().
>>>>>
>>>>> drivers/pci/pci.c
>>>>> static void do_pci_disable_device(struct pci_dev *dev)
>>>>> {
>>>>>           u16 pci_command;
>>>>>
>>>>>           pci_read_config_word(dev, PCI_COMMAND, &pci_command);
>>>>>           if (pci_command & PCI_COMMAND_MASTER) {
>>>>>                   pci_command &= ~PCI_COMMAND_MASTER;
>>>>>                   pci_write_config_word(dev, PCI_COMMAND, 
>>>>> pci_command);
>>>>>           }
>>>>>
>>>>>           pcibios_disable_device(dev);
>>>>> }
>>>>>
>>>>> When remove "pci_command &= ~PCI_COMMAND_MASTER;", it can work 
>>>>> well when
>>>>> shutdown or reboot.
>>>>>
>>>>> As Oliver O'Halloran said, no need to call pci_disable_device() when
>>>>> actually shutting down, but we should call pci_disable_device() 
>>>>> before
>>>>> handing over to the new kernel on kexec reboot, so we can do some
>>>>> condition checks which are already executed afterwards by the 
>>>>> function
>>>>> pci_device_shutdown(), this is done by commit 4fc9bbf98fd6 ("PCI: 
>>>>> Disable
>>>>> Bus Master only on kexec reboot") and commit 6e0eda3c3898 ("PCI: 
>>>>> Don't try
>>>>> to disable Bus Master on disconnected PCI devices").
>>>>>
>>>>> drivers/pci/pci-driver.c
>>>>> static void pci_device_shutdown(struct device *dev)
>>>>> {
>>>>>    ...
>>>>>           if (drv && drv->shutdown)
>>>>>                   drv->shutdown(pci_dev);
>>>>>
>>>>>           /*
>>>>>            * If this is a kexec reboot, turn off Bus Master bit on 
>>>>> the
>>>>>            * device to tell it to not continue to do DMA. Don't touch
>>>>>            * devices in D3cold or unknown states.
>>>>>            * If it is not a kexec reboot, firmware will hit the PCI
>>>>>            * devices with big hammer and stop their DMA any way.
>>>>>            */
>>>>>           if (kexec_in_progress && (pci_dev->current_state <= 
>>>>> PCI_D3hot))
>>>>>                   pci_clear_master(pci_dev);
>>>>> }
>>>>>
>>>>> [   36.159446] Call Trace:
>>>>> [   36.241688] [<ffffffff80211434>] show_stack+0x9c/0x130
>>>>> [   36.326619] [<ffffffff80661b70>] dump_stack+0xb0/0xf0
>>>>> [   36.410403] [<ffffffff806a8240>] pcie_portdrv_shutdown+0x18/0x78
>>>>> [   36.495302] [<ffffffff8069c6b4>] pci_device_shutdown+0x44/0x90
>>>>> [   36.580027] [<ffffffff807aac90>] device_shutdown+0x130/0x290
>>>>> [   36.664486] [<ffffffff80265448>] kernel_power_off+0x38/0x80
>>>>> [   36.748272] [<ffffffff80265634>] __do_sys_reboot+0x1a4/0x258
>>>>> [   36.831985] [<ffffffff80218b90>] syscall_common+0x34/0x58
>>>>>
>>>>> Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
>>>>> ---
>>>>>    drivers/pci/pcie/portdrv_core.c |  1 -
>>>>>    drivers/pci/pcie/portdrv_pci.c  | 14 +++++++++++++-
>>>>>    2 files changed, 13 insertions(+), 2 deletions(-)
>>>>>
>>>>> diff --git a/drivers/pci/pcie/portdrv_core.c 
>>>>> b/drivers/pci/pcie/portdrv_core.c
>>>>> index 50a9522..1991aca 100644
>>>>> --- a/drivers/pci/pcie/portdrv_core.c
>>>>> +++ b/drivers/pci/pcie/portdrv_core.c
>>>>> @@ -491,7 +491,6 @@ void pcie_port_device_remove(struct pci_dev *dev)
>>>>>    {
>>>>>           device_for_each_child(&dev->dev, NULL, remove_iter);
>>>>>           pci_free_irq_vectors(dev);
>>>>> -       pci_disable_device(dev);
>>>>>    }
>>>>>
>>>>>    /**
>>>>> diff --git a/drivers/pci/pcie/portdrv_pci.c 
>>>>> b/drivers/pci/pcie/portdrv_pci.c
>>>>> index 3a3ce40..cab37a8 100644
>>>>> --- a/drivers/pci/pcie/portdrv_pci.c
>>>>> +++ b/drivers/pci/pcie/portdrv_pci.c
>>>>> @@ -143,6 +143,18 @@ static void pcie_portdrv_remove(struct 
>>>>> pci_dev *dev)
>>>>>           }
>>>>>
>>>>>           pcie_port_device_remove(dev);
>>>>> +       pci_disable_device(dev);
>>>>> +}
>>>>> +
>>>>> +static void pcie_portdrv_shutdown(struct pci_dev *dev)
>>>>> +{
>>>>> +       if (pci_bridge_d3_possible(dev)) {
>>>>> +               pm_runtime_forbid(&dev->dev);
>>>>> +               pm_runtime_get_noresume(&dev->dev);
>>>>> + pm_runtime_dont_use_autosuspend(&dev->dev);
>>>>> +       }
>>>>> +
>>>>> +       pcie_port_device_remove(dev);
>>>>>    }
>>>>>
>>>>>    static pci_ers_result_t pcie_portdrv_error_detected(struct 
>>>>> pci_dev *dev,
>>>>> @@ -211,7 +223,7 @@ static struct pci_driver pcie_portdriver = {
>>>>>
>>>>>           .probe          = pcie_portdrv_probe,
>>>>>           .remove         = pcie_portdrv_remove,
>>>>> -       .shutdown       = pcie_portdrv_remove,
>>>>> +       .shutdown       = pcie_portdrv_shutdown,
>>>>>
>>>>>           .err_handler    = &pcie_portdrv_err_handler,
>>>>>
>>>>> -- 
>>>>> 2.1.0
Tiezhu Yang Sept. 14, 2020, 12:33 p.m. UTC | #7
On 09/14/2020 08:06 PM, Zhou Yanjie wrote:
> Hello Tiezhu
>
> 在 2020/9/14 下午7:24, Tiezhu Yang 写道:
>> On 09/14/2020 05:46 PM, Huacai Chen wrote:
>>> Hi, Tiezhu,
>>>
>>> On Mon, Sep 14, 2020 at 5:30 PM Tiezhu Yang <yangtiezhu@loongson.cn> 
>>> wrote:
>>>> On 09/14/2020 04:52 PM, Huacai Chen wrote:
>>>>> Hi, Tiezhu,
>>>>>
>>>>> How do you test kexec? kexec -e or systemctl kexec? Or both?
>>>> kexec -l vmlinux --append="root=/dev/sda2 console=ttyS0,115200"
>>>> kexec -e
>>> So you haven't tested "systemctl kexec"?
>>
>> Yes, the distro I used is Loongnix which has not kexec service now.
>> Is there any problem when use systemctl kexec? If you have more details,
>> please let me know.
>>
>
> It is inevitable that there will be some omissions during testing. 
> This is understandable, but what I am more curious about is why you 
> send the patch that Huacai has already sent every time. Why not 
> strengthen communication with Huacai and avoid duplication of work?

Oh, maybe there exists some misunderstanding, let me explain it.

when I update the latest kernel, the shutdown or reboot failed, I debug it
and try to solve it, then I sent a RFC patch.

I notice that there exists a similar patch by Huacai until Bjorn Helgaas cc
Huacai's patch to me. Because I have a different thought, so I sent RFC v2
and v3.

I think cc the related people at first can avoid this case.

>
>
>>>
>>> Huacai
>>>>> P.S., Please also CC my gmail (chenhuacai@gmail.com) since lemote.com
>>>>> has some communication problems.
>>>> OK, no problem.
>>>>
>>>>> Huacai
>>>>>
>>>>>> 陈华才江苏航天龙梦信息技术有限公司/研发中心/软件部 ------------------ 
>>>>>> Original ------------------From: "Tiezhu 
>>>>>> Yang"<yangtiezhu@loongson.cn>;Date:  Mon, Sep 14, 2020 03:57 
>>>>>> PMTo:  "Bjorn Helgaas"<bhelgaas@google.com>; Cc: 
>>>>>> "linux-pci"<linux-pci@vger.kernel.org>; 
>>>>>> "linux-kernel"<linux-kernel@vger.kernel.org>; "Rafael J. 
>>>>>> Wysocki"<rafael.j.wysocki@intel.com>; "Konstantin 
>>>>>> Khlebnikov"<khlebnikov@openvz.org>; "Khalid 
>>>>>> Aziz"<khalid.aziz@oracle.com>; "Vivek Goyal"<vgoyal@redhat.com>; 
>>>>>> "Lukas Wunner"<lukas@wunner.de>; "Oliver 
>>>>>> O'Halloran"<oohall@gmail.com>; "Huacai Chen"<chenhc@lemote.com>; 
>>>>>> "Jiaxun Yang"<jiaxun.yang@flygoat.com>; "Xuefeng 
>>>>>> Li"<lixuefeng@loongson.cn>; Subject:  [RFC PATCH v3] PCI/portdrv: 
>>>>>> Only disable Bus Master on kexec reboot and connected PCI devices 
>>>>>> After commit 745be2e700cd ("PCIe: portdrv: call pci_disable_device
>>>>>> during remove") and commit cc27b735ad3a ("PCI/portdrv: Turn off PCIe
>>>>>> services during shutdown"), it also calls pci_disable_device() 
>>>>>> during
>>>>>> shutdown, this leads to shutdown or reboot failure occasionally 
>>>>>> due to
>>>>>> clear PCI_COMMAND_MASTER on the device in do_pci_disable_device().
>>>>>>
>>>>>> drivers/pci/pci.c
>>>>>> static void do_pci_disable_device(struct pci_dev *dev)
>>>>>> {
>>>>>>           u16 pci_command;
>>>>>>
>>>>>>           pci_read_config_word(dev, PCI_COMMAND, &pci_command);
>>>>>>           if (pci_command & PCI_COMMAND_MASTER) {
>>>>>>                   pci_command &= ~PCI_COMMAND_MASTER;
>>>>>>                   pci_write_config_word(dev, PCI_COMMAND, 
>>>>>> pci_command);
>>>>>>           }
>>>>>>
>>>>>>           pcibios_disable_device(dev);
>>>>>> }
>>>>>>
>>>>>> When remove "pci_command &= ~PCI_COMMAND_MASTER;", it can work 
>>>>>> well when
>>>>>> shutdown or reboot.
>>>>>>
>>>>>> As Oliver O'Halloran said, no need to call pci_disable_device() when
>>>>>> actually shutting down, but we should call pci_disable_device() 
>>>>>> before
>>>>>> handing over to the new kernel on kexec reboot, so we can do some
>>>>>> condition checks which are already executed afterwards by the 
>>>>>> function
>>>>>> pci_device_shutdown(), this is done by commit 4fc9bbf98fd6 ("PCI: 
>>>>>> Disable
>>>>>> Bus Master only on kexec reboot") and commit 6e0eda3c3898 ("PCI: 
>>>>>> Don't try
>>>>>> to disable Bus Master on disconnected PCI devices").
>>>>>>
>>>>>> drivers/pci/pci-driver.c
>>>>>> static void pci_device_shutdown(struct device *dev)
>>>>>> {
>>>>>>    ...
>>>>>>           if (drv && drv->shutdown)
>>>>>>                   drv->shutdown(pci_dev);
>>>>>>
>>>>>>           /*
>>>>>>            * If this is a kexec reboot, turn off Bus Master bit 
>>>>>> on the
>>>>>>            * device to tell it to not continue to do DMA. Don't 
>>>>>> touch
>>>>>>            * devices in D3cold or unknown states.
>>>>>>            * If it is not a kexec reboot, firmware will hit the PCI
>>>>>>            * devices with big hammer and stop their DMA any way.
>>>>>>            */
>>>>>>           if (kexec_in_progress && (pci_dev->current_state <= 
>>>>>> PCI_D3hot))
>>>>>>                   pci_clear_master(pci_dev);
>>>>>> }
>>>>>>
>>>>>> [   36.159446] Call Trace:
>>>>>> [   36.241688] [<ffffffff80211434>] show_stack+0x9c/0x130
>>>>>> [   36.326619] [<ffffffff80661b70>] dump_stack+0xb0/0xf0
>>>>>> [   36.410403] [<ffffffff806a8240>] pcie_portdrv_shutdown+0x18/0x78
>>>>>> [   36.495302] [<ffffffff8069c6b4>] pci_device_shutdown+0x44/0x90
>>>>>> [   36.580027] [<ffffffff807aac90>] device_shutdown+0x130/0x290
>>>>>> [   36.664486] [<ffffffff80265448>] kernel_power_off+0x38/0x80
>>>>>> [   36.748272] [<ffffffff80265634>] __do_sys_reboot+0x1a4/0x258
>>>>>> [   36.831985] [<ffffffff80218b90>] syscall_common+0x34/0x58
>>>>>>
>>>>>> Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
>>>>>> ---
>>>>>>    drivers/pci/pcie/portdrv_core.c |  1 -
>>>>>>    drivers/pci/pcie/portdrv_pci.c  | 14 +++++++++++++-
>>>>>>    2 files changed, 13 insertions(+), 2 deletions(-)
>>>>>>
>>>>>> diff --git a/drivers/pci/pcie/portdrv_core.c 
>>>>>> b/drivers/pci/pcie/portdrv_core.c
>>>>>> index 50a9522..1991aca 100644
>>>>>> --- a/drivers/pci/pcie/portdrv_core.c
>>>>>> +++ b/drivers/pci/pcie/portdrv_core.c
>>>>>> @@ -491,7 +491,6 @@ void pcie_port_device_remove(struct pci_dev 
>>>>>> *dev)
>>>>>>    {
>>>>>>           device_for_each_child(&dev->dev, NULL, remove_iter);
>>>>>>           pci_free_irq_vectors(dev);
>>>>>> -       pci_disable_device(dev);
>>>>>>    }
>>>>>>
>>>>>>    /**
>>>>>> diff --git a/drivers/pci/pcie/portdrv_pci.c 
>>>>>> b/drivers/pci/pcie/portdrv_pci.c
>>>>>> index 3a3ce40..cab37a8 100644
>>>>>> --- a/drivers/pci/pcie/portdrv_pci.c
>>>>>> +++ b/drivers/pci/pcie/portdrv_pci.c
>>>>>> @@ -143,6 +143,18 @@ static void pcie_portdrv_remove(struct 
>>>>>> pci_dev *dev)
>>>>>>           }
>>>>>>
>>>>>>           pcie_port_device_remove(dev);
>>>>>> +       pci_disable_device(dev);
>>>>>> +}
>>>>>> +
>>>>>> +static void pcie_portdrv_shutdown(struct pci_dev *dev)
>>>>>> +{
>>>>>> +       if (pci_bridge_d3_possible(dev)) {
>>>>>> +               pm_runtime_forbid(&dev->dev);
>>>>>> + pm_runtime_get_noresume(&dev->dev);
>>>>>> + pm_runtime_dont_use_autosuspend(&dev->dev);
>>>>>> +       }
>>>>>> +
>>>>>> +       pcie_port_device_remove(dev);
>>>>>>    }
>>>>>>
>>>>>>    static pci_ers_result_t pcie_portdrv_error_detected(struct 
>>>>>> pci_dev *dev,
>>>>>> @@ -211,7 +223,7 @@ static struct pci_driver pcie_portdriver = {
>>>>>>
>>>>>>           .probe          = pcie_portdrv_probe,
>>>>>>           .remove         = pcie_portdrv_remove,
>>>>>> -       .shutdown       = pcie_portdrv_remove,
>>>>>> +       .shutdown       = pcie_portdrv_shutdown,
>>>>>>
>>>>>>           .err_handler    = &pcie_portdrv_err_handler,
>>>>>>
>>>>>> -- 
>>>>>> 2.1.0
Huacai Chen Sept. 15, 2020, 1:36 a.m. UTC | #8
Hi, Tiezhu,

On Mon, Sep 14, 2020 at 7:25 PM Tiezhu Yang <yangtiezhu@loongson.cn> wrote:
>
> On 09/14/2020 05:46 PM, Huacai Chen wrote:
> > Hi, Tiezhu,
> >
> > On Mon, Sep 14, 2020 at 5:30 PM Tiezhu Yang <yangtiezhu@loongson.cn> wrote:
> >> On 09/14/2020 04:52 PM, Huacai Chen wrote:
> >>> Hi, Tiezhu,
> >>>
> >>> How do you test kexec? kexec -e or systemctl kexec? Or both?
> >> kexec -l vmlinux --append="root=/dev/sda2 console=ttyS0,115200"
> >> kexec -e
> > So you haven't tested "systemctl kexec"?
>
> Yes, the distro I used is Loongnix which has not kexec service now.
> Is there any problem when use systemctl kexec? If you have more details,
> please let me know.
If you use systemctl kexec, the first part of kexec is the same as a
normal reboot/poweroff. So, there is no reason that reboot/poweroff is
bad but kexec is good.
Then, Bjorn, what is the best solution now? It seems like my first
version is OK (commit messages should be improved, of course).

Huacai
>
> >
> > Huacai
> >>> P.S., Please also CC my gmail (chenhuacai@gmail.com) since lemote.com
> >>> has some communication problems.
> >> OK, no problem.
> >>
> >>> Huacai
> >>>
> >>>> 陈华才江苏航天龙梦信息技术有限公司/研发中心/软件部   ------------------ Original ------------------From:  "Tiezhu Yang"<yangtiezhu@loongson.cn>;Date:  Mon, Sep 14, 2020 03:57 PMTo:  "Bjorn Helgaas"<bhelgaas@google.com>; Cc:  "linux-pci"<linux-pci@vger.kernel.org>; "linux-kernel"<linux-kernel@vger.kernel.org>; "Rafael J. Wysocki"<rafael.j.wysocki@intel.com>; "Konstantin Khlebnikov"<khlebnikov@openvz.org>; "Khalid Aziz"<khalid.aziz@oracle.com>; "Vivek Goyal"<vgoyal@redhat.com>; "Lukas Wunner"<lukas@wunner.de>; "Oliver O'Halloran"<oohall@gmail.com>; "Huacai Chen"<chenhc@lemote.com>; "Jiaxun Yang"<jiaxun.yang@flygoat.com>; "Xuefeng Li"<lixuefeng@loongson.cn>; Subject:  [RFC PATCH v3] PCI/portdrv: Only disable Bus Master on kexec reboot and connected PCI devices After commit 745be2e700cd ("PCIe: portdrv: call pci_disable_device
> >>>> during remove") and commit cc27b735ad3a ("PCI/portdrv: Turn off PCIe
> >>>> services during shutdown"), it also calls pci_disable_device() during
> >>>> shutdown, this leads to shutdown or reboot failure occasionally due to
> >>>> clear PCI_COMMAND_MASTER on the device in do_pci_disable_device().
> >>>>
> >>>> drivers/pci/pci.c
> >>>> static void do_pci_disable_device(struct pci_dev *dev)
> >>>> {
> >>>>           u16 pci_command;
> >>>>
> >>>>           pci_read_config_word(dev, PCI_COMMAND, &pci_command);
> >>>>           if (pci_command & PCI_COMMAND_MASTER) {
> >>>>                   pci_command &= ~PCI_COMMAND_MASTER;
> >>>>                   pci_write_config_word(dev, PCI_COMMAND, pci_command);
> >>>>           }
> >>>>
> >>>>           pcibios_disable_device(dev);
> >>>> }
> >>>>
> >>>> When remove "pci_command &= ~PCI_COMMAND_MASTER;", it can work well when
> >>>> shutdown or reboot.
> >>>>
> >>>> As Oliver O'Halloran said, no need to call pci_disable_device() when
> >>>> actually shutting down, but we should call pci_disable_device() before
> >>>> handing over to the new kernel on kexec reboot, so we can do some
> >>>> condition checks which are already executed afterwards by the function
> >>>> pci_device_shutdown(), this is done by commit 4fc9bbf98fd6 ("PCI: Disable
> >>>> Bus Master only on kexec reboot") and commit 6e0eda3c3898 ("PCI: Don't try
> >>>> to disable Bus Master on disconnected PCI devices").
> >>>>
> >>>> drivers/pci/pci-driver.c
> >>>> static void pci_device_shutdown(struct device *dev)
> >>>> {
> >>>>    ...
> >>>>           if (drv && drv->shutdown)
> >>>>                   drv->shutdown(pci_dev);
> >>>>
> >>>>           /*
> >>>>            * If this is a kexec reboot, turn off Bus Master bit on the
> >>>>            * device to tell it to not continue to do DMA. Don't touch
> >>>>            * devices in D3cold or unknown states.
> >>>>            * If it is not a kexec reboot, firmware will hit the PCI
> >>>>            * devices with big hammer and stop their DMA any way.
> >>>>            */
> >>>>           if (kexec_in_progress && (pci_dev->current_state <= PCI_D3hot))
> >>>>                   pci_clear_master(pci_dev);
> >>>> }
> >>>>
> >>>> [   36.159446] Call Trace:
> >>>> [   36.241688] [<ffffffff80211434>] show_stack+0x9c/0x130
> >>>> [   36.326619] [<ffffffff80661b70>] dump_stack+0xb0/0xf0
> >>>> [   36.410403] [<ffffffff806a8240>] pcie_portdrv_shutdown+0x18/0x78
> >>>> [   36.495302] [<ffffffff8069c6b4>] pci_device_shutdown+0x44/0x90
> >>>> [   36.580027] [<ffffffff807aac90>] device_shutdown+0x130/0x290
> >>>> [   36.664486] [<ffffffff80265448>] kernel_power_off+0x38/0x80
> >>>> [   36.748272] [<ffffffff80265634>] __do_sys_reboot+0x1a4/0x258
> >>>> [   36.831985] [<ffffffff80218b90>] syscall_common+0x34/0x58
> >>>>
> >>>> Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
> >>>> ---
> >>>>    drivers/pci/pcie/portdrv_core.c |  1 -
> >>>>    drivers/pci/pcie/portdrv_pci.c  | 14 +++++++++++++-
> >>>>    2 files changed, 13 insertions(+), 2 deletions(-)
> >>>>
> >>>> diff --git a/drivers/pci/pcie/portdrv_core.c b/drivers/pci/pcie/portdrv_core.c
> >>>> index 50a9522..1991aca 100644
> >>>> --- a/drivers/pci/pcie/portdrv_core.c
> >>>> +++ b/drivers/pci/pcie/portdrv_core.c
> >>>> @@ -491,7 +491,6 @@ void pcie_port_device_remove(struct pci_dev *dev)
> >>>>    {
> >>>>           device_for_each_child(&dev->dev, NULL, remove_iter);
> >>>>           pci_free_irq_vectors(dev);
> >>>> -       pci_disable_device(dev);
> >>>>    }
> >>>>
> >>>>    /**
> >>>> diff --git a/drivers/pci/pcie/portdrv_pci.c b/drivers/pci/pcie/portdrv_pci.c
> >>>> index 3a3ce40..cab37a8 100644
> >>>> --- a/drivers/pci/pcie/portdrv_pci.c
> >>>> +++ b/drivers/pci/pcie/portdrv_pci.c
> >>>> @@ -143,6 +143,18 @@ static void pcie_portdrv_remove(struct pci_dev *dev)
> >>>>           }
> >>>>
> >>>>           pcie_port_device_remove(dev);
> >>>> +       pci_disable_device(dev);
> >>>> +}
> >>>> +
> >>>> +static void pcie_portdrv_shutdown(struct pci_dev *dev)
> >>>> +{
> >>>> +       if (pci_bridge_d3_possible(dev)) {
> >>>> +               pm_runtime_forbid(&dev->dev);
> >>>> +               pm_runtime_get_noresume(&dev->dev);
> >>>> +               pm_runtime_dont_use_autosuspend(&dev->dev);
> >>>> +       }
> >>>> +
> >>>> +       pcie_port_device_remove(dev);
> >>>>    }
> >>>>
> >>>>    static pci_ers_result_t pcie_portdrv_error_detected(struct pci_dev *dev,
> >>>> @@ -211,7 +223,7 @@ static struct pci_driver pcie_portdriver = {
> >>>>
> >>>>           .probe          = pcie_portdrv_probe,
> >>>>           .remove         = pcie_portdrv_remove,
> >>>> -       .shutdown       = pcie_portdrv_remove,
> >>>> +       .shutdown       = pcie_portdrv_shutdown,
> >>>>
> >>>>           .err_handler    = &pcie_portdrv_err_handler,
> >>>>
> >>>> --
> >>>> 2.1.0
>
Bjorn Helgaas Sept. 15, 2020, 8:38 p.m. UTC | #9
On Tue, Sep 15, 2020 at 09:36:13AM +0800, Huacai Chen wrote:
> Hi, Tiezhu,
> 
> On Mon, Sep 14, 2020 at 7:25 PM Tiezhu Yang <yangtiezhu@loongson.cn> wrote:
> >
> > On 09/14/2020 05:46 PM, Huacai Chen wrote:
> > > Hi, Tiezhu,
> > >
> > > On Mon, Sep 14, 2020 at 5:30 PM Tiezhu Yang <yangtiezhu@loongson.cn> wrote:
> > >> On 09/14/2020 04:52 PM, Huacai Chen wrote:
> > >>> Hi, Tiezhu,
> > >>>
> > >>> How do you test kexec? kexec -e or systemctl kexec? Or both?
> > >> kexec -l vmlinux --append="root=/dev/sda2 console=ttyS0,115200"
> > >> kexec -e
> > > So you haven't tested "systemctl kexec"?
> >
> > Yes, the distro I used is Loongnix which has not kexec service now.
> > Is there any problem when use systemctl kexec? If you have more details,
> > please let me know.
> If you use systemctl kexec, the first part of kexec is the same as a
> normal reboot/poweroff. So, there is no reason that reboot/poweroff is
> bad but kexec is good.
> Then, Bjorn, what is the best solution now? It seems like my first
> version is OK (commit messages should be improved, of course).

Sorry, there are too many conversations going on for me to keep track
of what first version you're referring to.  Please include a specific
lore.kernel.org URL.

Bjorn
Huacai Chen Sept. 16, 2020, 6:19 a.m. UTC | #10
Hi, Bjorn,

On Tue, Sep 15, 2020 at 4:38 PM Bjorn Helgaas <helgaas@kernel.org> wrote:
>
> On Tue, Sep 15, 2020 at 09:36:13AM +0800, Huacai Chen wrote:
> > Hi, Tiezhu,
> >
> > On Mon, Sep 14, 2020 at 7:25 PM Tiezhu Yang <yangtiezhu@loongson.cn> wrote:
> > >
> > > On 09/14/2020 05:46 PM, Huacai Chen wrote:
> > > > Hi, Tiezhu,
> > > >
> > > > On Mon, Sep 14, 2020 at 5:30 PM Tiezhu Yang <yangtiezhu@loongson.cn> wrote:
> > > >> On 09/14/2020 04:52 PM, Huacai Chen wrote:
> > > >>> Hi, Tiezhu,
> > > >>>
> > > >>> How do you test kexec? kexec -e or systemctl kexec? Or both?
> > > >> kexec -l vmlinux --append="root=/dev/sda2 console=ttyS0,115200"
> > > >> kexec -e
> > > > So you haven't tested "systemctl kexec"?
> > >
> > > Yes, the distro I used is Loongnix which has not kexec service now.
> > > Is there any problem when use systemctl kexec? If you have more details,
> > > please let me know.
> > If you use systemctl kexec, the first part of kexec is the same as a
> > normal reboot/poweroff. So, there is no reason that reboot/poweroff is
> > bad but kexec is good.
> > Then, Bjorn, what is the best solution now? It seems like my first
> > version is OK (commit messages should be improved, of course).
>
> Sorry, there are too many conversations going on for me to keep track
> of what first version you're referring to.  Please include a specific
> lore.kernel.org URL.
My first version means this one:
https://patchwork.kernel.org/patch/11695929/

Thanks,
Huacai
>
> Bjorn
diff mbox series

Patch

diff --git a/drivers/pci/pcie/portdrv_core.c b/drivers/pci/pcie/portdrv_core.c
index 50a9522..1991aca 100644
--- a/drivers/pci/pcie/portdrv_core.c
+++ b/drivers/pci/pcie/portdrv_core.c
@@ -491,7 +491,6 @@  void pcie_port_device_remove(struct pci_dev *dev)
 {
 	device_for_each_child(&dev->dev, NULL, remove_iter);
 	pci_free_irq_vectors(dev);
-	pci_disable_device(dev);
 }
 
 /**
diff --git a/drivers/pci/pcie/portdrv_pci.c b/drivers/pci/pcie/portdrv_pci.c
index 3a3ce40..cab37a8 100644
--- a/drivers/pci/pcie/portdrv_pci.c
+++ b/drivers/pci/pcie/portdrv_pci.c
@@ -143,6 +143,18 @@  static void pcie_portdrv_remove(struct pci_dev *dev)
 	}
 
 	pcie_port_device_remove(dev);
+	pci_disable_device(dev);
+}
+
+static void pcie_portdrv_shutdown(struct pci_dev *dev)
+{
+	if (pci_bridge_d3_possible(dev)) {
+		pm_runtime_forbid(&dev->dev);
+		pm_runtime_get_noresume(&dev->dev);
+		pm_runtime_dont_use_autosuspend(&dev->dev);
+	}
+
+	pcie_port_device_remove(dev);
 }
 
 static pci_ers_result_t pcie_portdrv_error_detected(struct pci_dev *dev,
@@ -211,7 +223,7 @@  static struct pci_driver pcie_portdriver = {
 
 	.probe		= pcie_portdrv_probe,
 	.remove		= pcie_portdrv_remove,
-	.shutdown	= pcie_portdrv_remove,
+	.shutdown	= pcie_portdrv_shutdown,
 
 	.err_handler	= &pcie_portdrv_err_handler,