diff mbox series

[V6,7/8] PCI: Add "pci=disable_10bit_tag=" parameter for peer-to-peer support

Message ID 1627038402-114183-8-git-send-email-liudongdong3@huawei.com (mailing list archive)
State Superseded
Delegated to: Bjorn Helgaas
Headers show
Series PCI: Enable 10-Bit tag support for PCIe devices | expand

Commit Message

Dongdong Liu July 23, 2021, 11:06 a.m. UTC 
PCIe spec 5.0 r1.0 section 2.2.6.2 says that if an Endpoint supports
sending Requests to other Endpoints (as opposed to host memory), the
Endpoint must not send 10-Bit Tag Requests to another given Endpoint
unless an implementation-specific mechanism determines that the Endpoint
supports 10-Bit Tag Completer capability. Add "pci=disable_10bit_tag="
parameter to disable 10-Bit Tag Requester if the peer device does not
support the 10-Bit Tag Completer. This will make P2P traffic safe.

Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
---
 Documentation/admin-guide/kernel-parameters.txt |  7 ++++
 drivers/pci/pci.c                               | 56 +++++++++++++++++++++++++
 drivers/pci/pci.h                               |  1 +
 drivers/pci/pcie/portdrv_pci.c                  | 13 +++---
 drivers/pci/probe.c                             |  9 ++--
 5 files changed, 78 insertions(+), 8 deletions(-)

Comments

Leon Romanovsky July 23, 2021, 11:32 a.m. UTC | #1 
On Fri, Jul 23, 2021 at 07:06:41PM +0800, Dongdong Liu wrote:
> PCIe spec 5.0 r1.0 section 2.2.6.2 says that if an Endpoint supports
> sending Requests to other Endpoints (as opposed to host memory), the
> Endpoint must not send 10-Bit Tag Requests to another given Endpoint
> unless an implementation-specific mechanism determines that the Endpoint
> supports 10-Bit Tag Completer capability. Add "pci=disable_10bit_tag="
> parameter to disable 10-Bit Tag Requester if the peer device does not
> support the 10-Bit Tag Completer. This will make P2P traffic safe.
> 
> Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
> ---
>  Documentation/admin-guide/kernel-parameters.txt |  7 ++++
>  drivers/pci/pci.c                               | 56 +++++++++++++++++++++++++
>  drivers/pci/pci.h                               |  1 +
>  drivers/pci/pcie/portdrv_pci.c                  | 13 +++---
>  drivers/pci/probe.c                             |  9 ++--
>  5 files changed, 78 insertions(+), 8 deletions(-)
> 
> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> index bdb2200..c2c4585 100644
> --- a/Documentation/admin-guide/kernel-parameters.txt
> +++ b/Documentation/admin-guide/kernel-parameters.txt
> @@ -4019,6 +4019,13 @@
>  				bridges without forcing it upstream. Note:
>  				this removes isolation between devices and
>  				may put more devices in an IOMMU group.
> +		disable_10bit_tag=<pci_dev>[; ...]
> +				  Specify one or more PCI devices (in the format
> +				  specified above) separated by semicolons.
> +				  Disable 10-Bit Tag Requester if the peer
> +				  device does not support the 10-Bit Tag
> +				  Completer.This will make P2P traffic safe.

I can't imagine more awkward user experience than such kernel parameter.

As a user, I will need to boot the system, hope for the best that system
works, write down all PCI device numbers, guess which one doesn't work
properly, update grub with new command line argument and reboot the
system. Any HW change and this dance should be repeated.

Thanks
Logan Gunthorpe July 23, 2021, 4:20 p.m. UTC | #2 
On 2021-07-23 5:32 a.m., Leon Romanovsky wrote:
> On Fri, Jul 23, 2021 at 07:06:41PM +0800, Dongdong Liu wrote:
>> PCIe spec 5.0 r1.0 section 2.2.6.2 says that if an Endpoint supports
>> sending Requests to other Endpoints (as opposed to host memory), the
>> Endpoint must not send 10-Bit Tag Requests to another given Endpoint
>> unless an implementation-specific mechanism determines that the Endpoint
>> supports 10-Bit Tag Completer capability. Add "pci=disable_10bit_tag="
>> parameter to disable 10-Bit Tag Requester if the peer device does not
>> support the 10-Bit Tag Completer. This will make P2P traffic safe.
>>
>> Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
>> ---
>>  Documentation/admin-guide/kernel-parameters.txt |  7 ++++
>>  drivers/pci/pci.c                               | 56 +++++++++++++++++++++++++
>>  drivers/pci/pci.h                               |  1 +
>>  drivers/pci/pcie/portdrv_pci.c                  | 13 +++---
>>  drivers/pci/probe.c                             |  9 ++--
>>  5 files changed, 78 insertions(+), 8 deletions(-)
>>
>> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
>> index bdb2200..c2c4585 100644
>> --- a/Documentation/admin-guide/kernel-parameters.txt
>> +++ b/Documentation/admin-guide/kernel-parameters.txt
>> @@ -4019,6 +4019,13 @@
>>  				bridges without forcing it upstream. Note:
>>  				this removes isolation between devices and
>>  				may put more devices in an IOMMU group.
>> +		disable_10bit_tag=<pci_dev>[; ...]
>> +				  Specify one or more PCI devices (in the format
>> +				  specified above) separated by semicolons.
>> +				  Disable 10-Bit Tag Requester if the peer
>> +				  device does not support the 10-Bit Tag
>> +				  Completer.This will make P2P traffic safe.
> 
> I can't imagine more awkward user experience than such kernel parameter.
> 
> As a user, I will need to boot the system, hope for the best that system
> works, write down all PCI device numbers, guess which one doesn't work
> properly, update grub with new command line argument and reboot the
> system. Any HW change and this dance should be repeated.

There are already two such PCI parameters with this pattern and they are
not that awkward. pci_dev may be specified with either vendor/device IDS
or with a path of BDFs (which protects against renumbering).

This flag is only useful in P2PDMA traffic, and if the user attempts
such a transfer, it prints a warning (see the next patch) with the exact
parameter that needs to be added to the command line.

This has worked well for disable_acs_redir and was used for
resource_alignment before that for quite some time. So save a better
suggestion I think this is more than acceptable.

Logan
kernel test robot July 23, 2021, 4:58 p.m. UTC | #3 
Hi Dongdong,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on pci/next]
[also build test WARNING on linuxtv-media/master linus/master v5.14-rc2 next-20210723]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Dongdong-Liu/PCI-Enable-10-Bit-tag-support-for-PCIe-devices/20210723-190930
base:   https://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci.git next
config: x86_64-randconfig-b001-20210723 (attached as .config)
compiler: clang version 13.0.0 (https://github.com/llvm/llvm-project 9625ca5b602616b2f5584e8a49ba93c52c141e40)
reproduce (this is a W=1 build):
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # install x86_64 cross compiling tool for clang build
        # apt-get install binutils-x86-64-linux-gnu
        # https://github.com/0day-ci/linux/commit/2ff0b803971a3df5815c96c5c4874f4eef64fa2f
        git remote add linux-review https://github.com/0day-ci/linux
        git fetch --no-tags linux-review Dongdong-Liu/PCI-Enable-10-Bit-tag-support-for-PCIe-devices/20210723-190930
        git checkout 2ff0b803971a3df5815c96c5c4874f4eef64fa2f
        # save the attached .config to linux build tree
        mkdir build_dir
        COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross O=build_dir ARCH=x86_64 SHELL=/bin/bash drivers/pci/

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>

All warnings (new ones prefixed by >>):

   drivers/pci/pci.c:6618:34: error: expected identifier
           pcie_capability_clear_word(dev, PCI_EXP_DEVCTL2,
                                           ^
   include/uapi/linux/pci_regs.h:657:26: note: expanded from macro 'PCI_EXP_DEVCTL2'
   #define PCI_EXP_DEVCTL2         40      /* Device Control 2 */
                                   ^
>> drivers/pci/pci.c:6618:2: warning: declaration specifier missing, defaulting to 'int'
           pcie_capability_clear_word(dev, PCI_EXP_DEVCTL2,
           ^
           int
   drivers/pci/pci.c:6618:28: error: this function declaration is not a prototype [-Werror,-Wstrict-prototypes]
           pcie_capability_clear_word(dev, PCI_EXP_DEVCTL2,
                                     ^
   drivers/pci/pci.c:6618:2: error: conflicting types for 'pcie_capability_clear_word'
           pcie_capability_clear_word(dev, PCI_EXP_DEVCTL2,
           ^
   include/linux/pci.h:1161:19: note: previous definition is here
   static inline int pcie_capability_clear_word(struct pci_dev *dev, int pos,
                     ^
   drivers/pci/pci.c:6621:2: error: expected parameter declarator
           pci_info(dev, "disabled 10-Bit Tag Requester\n");
           ^
   include/linux/pci.h:2472:46: note: expanded from macro 'pci_info'
   #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
                                                    ^
   drivers/pci/pci.c:6621:2: error: expected ')'
   include/linux/pci.h:2472:46: note: expanded from macro 'pci_info'
   #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
                                                    ^
   drivers/pci/pci.c:6621:2: note: to match this '('
   include/linux/pci.h:2472:37: note: expanded from macro 'pci_info'
   #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
                                           ^
   include/linux/dev_printk.h:118:11: note: expanded from macro 'dev_info'
           _dev_info(dev, dev_fmt(fmt), ##__VA_ARGS__)
                    ^
   drivers/pci/pci.c:6621:2: warning: declaration specifier missing, defaulting to 'int'
           pci_info(dev, "disabled 10-Bit Tag Requester\n");
           ^
           int
   include/linux/pci.h:2472:37: note: expanded from macro 'pci_info'
   #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
                                           ^
   include/linux/dev_printk.h:118:2: note: expanded from macro 'dev_info'
           _dev_info(dev, dev_fmt(fmt), ##__VA_ARGS__)
           ^
   drivers/pci/pci.c:6621:2: error: this function declaration is not a prototype [-Werror,-Wstrict-prototypes]
   include/linux/pci.h:2472:37: note: expanded from macro 'pci_info'
   #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
                                           ^
   include/linux/dev_printk.h:118:11: note: expanded from macro 'dev_info'
           _dev_info(dev, dev_fmt(fmt), ##__VA_ARGS__)
                    ^
   drivers/pci/pci.c:6621:2: error: conflicting types for '_dev_info'
   include/linux/pci.h:2472:37: note: expanded from macro 'pci_info'
   #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
                                           ^
   include/linux/dev_printk.h:118:2: note: expanded from macro 'dev_info'
           _dev_info(dev, dev_fmt(fmt), ##__VA_ARGS__)
           ^
   include/linux/dev_printk.h:56:6: note: previous declaration is here
   void _dev_info(const struct device *dev, const char *fmt, ...);
        ^
   drivers/pci/pci.c:6622:1: error: extraneous closing brace ('}')
   }
   ^
   2 warnings and 8 errors generated.


vim +/int +6618 drivers/pci/pci.c

  6580	
  6581		if (!disable_10bit_tag_param)
  6582			return;
  6583	
  6584		p = disable_10bit_tag_param;
  6585		while (*p) {
  6586			ret = pci_dev_str_match(dev, p, &p);
  6587			if (ret < 0) {
  6588				pr_info_once("PCI: Can't parse disable_10bit_tag parameter: %s\n",
  6589					     disable_10bit_tag_param);
  6590	
  6591				break;
  6592			} else if (ret == 1) {
  6593				/* Found a match */
  6594				break;
  6595			}
  6596	
  6597			if (*p != ';' && *p != ',') {
  6598				/* End of param or invalid format */
  6599				break;
  6600			}
  6601			p++;
  6602		}
  6603	
  6604		if (ret != 1)
  6605			return;
  6606	
  6607	#ifdef CONFIG_PCI_IOV
  6608		if (dev->is_virtfn) {
  6609			iov = dev->physfn->sriov;
  6610			iov->ctrl &= ~PCI_SRIOV_CTRL_VF_10BIT_TAG_REQ_EN;
  6611			pci_write_config_word(dev, iov->pos + PCI_SRIOV_CTRL,
  6612					      iov->ctrl);
  6613			pci_info(dev, "disabled PF SRIOV 10-Bit Tag Requester\n");
  6614			return;
  6615	#endif
  6616		}
  6617	
> 6618		pcie_capability_clear_word(dev, PCI_EXP_DEVCTL2,
  6619					   PCI_EXP_DEVCTL2_10BIT_TAG_REQ_EN);
  6620	
  6621		pci_info(dev, "disabled 10-Bit Tag Requester\n");
  6622	}
  6623	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org
Dongdong Liu July 24, 2021, 10:35 a.m. UTC | #4 
On 2021/7/24 0:58, kernel test robot wrote:
> Hi Dongdong,
>
> Thank you for the patch! Perhaps something to improve:
>
> [auto build test WARNING on pci/next]
> [also build test WARNING on linuxtv-media/master linus/master v5.14-rc2 next-20210723]
> [If your patch is applied to the wrong git tree, kindly drop us a note.
> And when submitting patch, we suggest to use '--base' as documented in
> https://git-scm.com/docs/git-format-patch]
>
> url:    https://github.com/0day-ci/linux/commits/Dongdong-Liu/PCI-Enable-10-Bit-tag-support-for-PCIe-devices/20210723-190930
> base:   https://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci.git next
> config: x86_64-randconfig-b001-20210723 (attached as .config)
> compiler: clang version 13.0.0 (https://github.com/llvm/llvm-project 9625ca5b602616b2f5584e8a49ba93c52c141e40)
> reproduce (this is a W=1 build):
>         wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
>         chmod +x ~/bin/make.cross
>         # install x86_64 cross compiling tool for clang build
>         # apt-get install binutils-x86-64-linux-gnu
>         # https://github.com/0day-ci/linux/commit/2ff0b803971a3df5815c96c5c4874f4eef64fa2f
>         git remote add linux-review https://github.com/0day-ci/linux
>         git fetch --no-tags linux-review Dongdong-Liu/PCI-Enable-10-Bit-tag-support-for-PCIe-devices/20210723-190930
>         git checkout 2ff0b803971a3df5815c96c5c4874f4eef64fa2f
>         # save the attached .config to linux build tree
>         mkdir build_dir
>         COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross O=build_dir ARCH=x86_64 SHELL=/bin/bash drivers/pci/
>
> If you fix the issue, kindly add following tag as appropriate
> Reported-by: kernel test robot <lkp@intel.com>
>
> All warnings (new ones prefixed by >>):
>
>    drivers/pci/pci.c:6618:34: error: expected identifier
>            pcie_capability_clear_word(dev, PCI_EXP_DEVCTL2,
>                                            ^
>    include/uapi/linux/pci_regs.h:657:26: note: expanded from macro 'PCI_EXP_DEVCTL2'
>    #define PCI_EXP_DEVCTL2         40      /* Device Control 2 */
>                                    ^
>>> drivers/pci/pci.c:6618:2: warning: declaration specifier missing, defaulting to 'int'
>            pcie_capability_clear_word(dev, PCI_EXP_DEVCTL2,
>            ^
>            int
>    drivers/pci/pci.c:6618:28: error: this function declaration is not a prototype [-Werror,-Wstrict-prototypes]
>            pcie_capability_clear_word(dev, PCI_EXP_DEVCTL2,
>                                      ^
>    drivers/pci/pci.c:6618:2: error: conflicting types for 'pcie_capability_clear_word'
>            pcie_capability_clear_word(dev, PCI_EXP_DEVCTL2,
>            ^
>    include/linux/pci.h:1161:19: note: previous definition is here
>    static inline int pcie_capability_clear_word(struct pci_dev *dev, int pos,
>                      ^
>    drivers/pci/pci.c:6621:2: error: expected parameter declarator
>            pci_info(dev, "disabled 10-Bit Tag Requester\n");
>            ^
>    include/linux/pci.h:2472:46: note: expanded from macro 'pci_info'
>    #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
>                                                     ^
>    drivers/pci/pci.c:6621:2: error: expected ')'
>    include/linux/pci.h:2472:46: note: expanded from macro 'pci_info'
>    #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
>                                                     ^
>    drivers/pci/pci.c:6621:2: note: to match this '('
>    include/linux/pci.h:2472:37: note: expanded from macro 'pci_info'
>    #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
>                                            ^
>    include/linux/dev_printk.h:118:11: note: expanded from macro 'dev_info'
>            _dev_info(dev, dev_fmt(fmt), ##__VA_ARGS__)
>                     ^
>    drivers/pci/pci.c:6621:2: warning: declaration specifier missing, defaulting to 'int'
>            pci_info(dev, "disabled 10-Bit Tag Requester\n");
>            ^
>            int
>    include/linux/pci.h:2472:37: note: expanded from macro 'pci_info'
>    #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
>                                            ^
>    include/linux/dev_printk.h:118:2: note: expanded from macro 'dev_info'
>            _dev_info(dev, dev_fmt(fmt), ##__VA_ARGS__)
>            ^
>    drivers/pci/pci.c:6621:2: error: this function declaration is not a prototype [-Werror,-Wstrict-prototypes]
>    include/linux/pci.h:2472:37: note: expanded from macro 'pci_info'
>    #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
>                                            ^
>    include/linux/dev_printk.h:118:11: note: expanded from macro 'dev_info'
>            _dev_info(dev, dev_fmt(fmt), ##__VA_ARGS__)
>                     ^
>    drivers/pci/pci.c:6621:2: error: conflicting types for '_dev_info'
>    include/linux/pci.h:2472:37: note: expanded from macro 'pci_info'
>    #define pci_info(pdev, fmt, arg...)     dev_info(&(pdev)->dev, fmt, ##arg)
>                                            ^
>    include/linux/dev_printk.h:118:2: note: expanded from macro 'dev_info'
>            _dev_info(dev, dev_fmt(fmt), ##__VA_ARGS__)
>            ^
>    include/linux/dev_printk.h:56:6: note: previous declaration is here
>    void _dev_info(const struct device *dev, const char *fmt, ...);
>         ^
>    drivers/pci/pci.c:6622:1: error: extraneous closing brace ('}')
>    }
>    ^
>    2 warnings and 8 errors generated.
>
>
> vim +/int +6618 drivers/pci/pci.c
>
>   6580	
>   6581		if (!disable_10bit_tag_param)
>   6582			return;
>   6583	
>   6584		p = disable_10bit_tag_param;
>   6585		while (*p) {
>   6586			ret = pci_dev_str_match(dev, p, &p);
>   6587			if (ret < 0) {
>   6588				pr_info_once("PCI: Can't parse disable_10bit_tag parameter: %s\n",
>   6589					     disable_10bit_tag_param);
>   6590	
>   6591				break;
>   6592			} else if (ret == 1) {
>   6593				/* Found a match */
>   6594				break;
>   6595			}
>   6596	
>   6597			if (*p != ';' && *p != ',') {
>   6598				/* End of param or invalid format */
>   6599				break;
>   6600			}
>   6601			p++;
>   6602		}
>   6603	
>   6604		if (ret != 1)
>   6605			return;
>   6606	
>   6607	#ifdef CONFIG_PCI_IOV
>   6608		if (dev->is_virtfn) {
>   6609			iov = dev->physfn->sriov;
>   6610			iov->ctrl &= ~PCI_SRIOV_CTRL_VF_10BIT_TAG_REQ_EN;
>   6611			pci_write_config_word(dev, iov->pos + PCI_SRIOV_CTRL,
>   6612					      iov->ctrl);
>   6613			pci_info(dev, "disabled PF SRIOV 10-Bit Tag Requester\n");
>   6614			return;
>   6615	#endif
>   6616		}
I made a mistake here, will fix.

Thanks,
Dongdong
>   6617	
>> 6618		pcie_capability_clear_word(dev, PCI_EXP_DEVCTL2,
>   6619					   PCI_EXP_DEVCTL2_10BIT_TAG_REQ_EN);
>   6620	
>   6621		pci_info(dev, "disabled 10-Bit Tag Requester\n");
>   6622	}
>   6623	
>
> ---
> 0-DAY CI Kernel Test Service, Intel Corporation
> https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org
>
Leon Romanovsky July 25, 2021, 6:39 a.m. UTC | #5 
On Fri, Jul 23, 2021 at 10:20:50AM -0600, Logan Gunthorpe wrote:
> 
> 
> 
> On 2021-07-23 5:32 a.m., Leon Romanovsky wrote:
> > On Fri, Jul 23, 2021 at 07:06:41PM +0800, Dongdong Liu wrote:
> >> PCIe spec 5.0 r1.0 section 2.2.6.2 says that if an Endpoint supports
> >> sending Requests to other Endpoints (as opposed to host memory), the
> >> Endpoint must not send 10-Bit Tag Requests to another given Endpoint
> >> unless an implementation-specific mechanism determines that the Endpoint
> >> supports 10-Bit Tag Completer capability. Add "pci=disable_10bit_tag="
> >> parameter to disable 10-Bit Tag Requester if the peer device does not
> >> support the 10-Bit Tag Completer. This will make P2P traffic safe.
> >>
> >> Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
> >> ---
> >>  Documentation/admin-guide/kernel-parameters.txt |  7 ++++
> >>  drivers/pci/pci.c                               | 56 +++++++++++++++++++++++++
> >>  drivers/pci/pci.h                               |  1 +
> >>  drivers/pci/pcie/portdrv_pci.c                  | 13 +++---
> >>  drivers/pci/probe.c                             |  9 ++--
> >>  5 files changed, 78 insertions(+), 8 deletions(-)
> >>
> >> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> >> index bdb2200..c2c4585 100644
> >> --- a/Documentation/admin-guide/kernel-parameters.txt
> >> +++ b/Documentation/admin-guide/kernel-parameters.txt
> >> @@ -4019,6 +4019,13 @@
> >>  				bridges without forcing it upstream. Note:
> >>  				this removes isolation between devices and
> >>  				may put more devices in an IOMMU group.
> >> +		disable_10bit_tag=<pci_dev>[; ...]
> >> +				  Specify one or more PCI devices (in the format
> >> +				  specified above) separated by semicolons.
> >> +				  Disable 10-Bit Tag Requester if the peer
> >> +				  device does not support the 10-Bit Tag
> >> +				  Completer.This will make P2P traffic safe.
> > 
> > I can't imagine more awkward user experience than such kernel parameter.
> > 
> > As a user, I will need to boot the system, hope for the best that system
> > works, write down all PCI device numbers, guess which one doesn't work
> > properly, update grub with new command line argument and reboot the
> > system. Any HW change and this dance should be repeated.
> 
> There are already two such PCI parameters with this pattern and they are
> not that awkward. pci_dev may be specified with either vendor/device IDS
> or with a path of BDFs (which protects against renumbering).

Unfortunately, in the real world, BDF is not so stable. It changes with
addition of new hardware, BIOS upgrades and even broken servers.

Vendor/device IDs doesn't work if you have multiple devices of same
vendor in the system.

> 
> This flag is only useful in P2PDMA traffic, and if the user attempts
> such a transfer, it prints a warning (see the next patch) with the exact
> parameter that needs to be added to the command line.

Dongdong citied PCI spec and it was very clear - don't enable this
feature unless you clearly know that it is safe to enable. This is
completely opposite to the proposal here - always enable and disable
if something is printed to the dmesg.

> 
> This has worked well for disable_acs_redir and was used for
> resource_alignment before that for quite some time. So save a better
> suggestion I think this is more than acceptable.

I don't know about other parameters and their history, but we are not in
90s anymore and addition of modules parameters (for the PCI it is kernel
cmdline arguments) are better to be changed to some configuration tool/sysfs.

Even FW upgrade with such kernel parameter can be problematic.

Thanks

> 
> Logan
Logan Gunthorpe July 26, 2021, 3:48 p.m. UTC | #6 
On 2021-07-25 12:39 a.m., Leon Romanovsky wrote:
> On Fri, Jul 23, 2021 at 10:20:50AM -0600, Logan Gunthorpe wrote:
>>
>>
>>
>> On 2021-07-23 5:32 a.m., Leon Romanovsky wrote:
>>> On Fri, Jul 23, 2021 at 07:06:41PM +0800, Dongdong Liu wrote:
>>>> PCIe spec 5.0 r1.0 section 2.2.6.2 says that if an Endpoint supports
>>>> sending Requests to other Endpoints (as opposed to host memory), the
>>>> Endpoint must not send 10-Bit Tag Requests to another given Endpoint
>>>> unless an implementation-specific mechanism determines that the Endpoint
>>>> supports 10-Bit Tag Completer capability. Add "pci=disable_10bit_tag="
>>>> parameter to disable 10-Bit Tag Requester if the peer device does not
>>>> support the 10-Bit Tag Completer. This will make P2P traffic safe.
>>>>
>>>> Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
>>>> ---
>>>>  Documentation/admin-guide/kernel-parameters.txt |  7 ++++
>>>>  drivers/pci/pci.c                               | 56 +++++++++++++++++++++++++
>>>>  drivers/pci/pci.h                               |  1 +
>>>>  drivers/pci/pcie/portdrv_pci.c                  | 13 +++---
>>>>  drivers/pci/probe.c                             |  9 ++--
>>>>  5 files changed, 78 insertions(+), 8 deletions(-)
>>>>
>>>> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
>>>> index bdb2200..c2c4585 100644
>>>> --- a/Documentation/admin-guide/kernel-parameters.txt
>>>> +++ b/Documentation/admin-guide/kernel-parameters.txt
>>>> @@ -4019,6 +4019,13 @@
>>>>  				bridges without forcing it upstream. Note:
>>>>  				this removes isolation between devices and
>>>>  				may put more devices in an IOMMU group.
>>>> +		disable_10bit_tag=<pci_dev>[; ...]
>>>> +				  Specify one or more PCI devices (in the format
>>>> +				  specified above) separated by semicolons.
>>>> +				  Disable 10-Bit Tag Requester if the peer
>>>> +				  device does not support the 10-Bit Tag
>>>> +				  Completer.This will make P2P traffic safe.
>>>
>>> I can't imagine more awkward user experience than such kernel parameter.
>>>
>>> As a user, I will need to boot the system, hope for the best that system
>>> works, write down all PCI device numbers, guess which one doesn't work
>>> properly, update grub with new command line argument and reboot the
>>> system. Any HW change and this dance should be repeated.
>>
>> There are already two such PCI parameters with this pattern and they are
>> not that awkward. pci_dev may be specified with either vendor/device IDS
>> or with a path of BDFs (which protects against renumbering).
> 
> Unfortunately, in the real world, BDF is not so stable. It changes with
> addition of new hardware, BIOS upgrades and even broken servers.

That's why it supports using a *path* of BDFs which tends not to catch
the wrong device if the topology changes.

> Vendor/device IDs doesn't work if you have multiple devices of same
> vendor in the system.

Yes, but it's fine for some use cases. That's why there's a range of
options.

>>
>> This flag is only useful in P2PDMA traffic, and if the user attempts
>> such a transfer, it prints a warning (see the next patch) with the exact
>> parameter that needs to be added to the command line.
> 
> Dongdong citied PCI spec and it was very clear - don't enable this
> feature unless you clearly know that it is safe to enable. This is
> completely opposite to the proposal here - always enable and disable
> if something is printed to the dmesg.

Quoting from patch 4:

"For platforms where the RC supports 10-Bit Tag Completer capability,
it is highly recommended for platform firmware or operating software
that configures PCIe hierarchies to Set the 10-Bit Tag Requester Enable
bit automatically in Endpoints with 10-Bit Tag Requester capability.
This enables the important class of 10-Bit Tag capable adapters that
send Memory Read Requests only to host memory."

Notice the last sentence. It's saying that devices who only talk to host
memory should have 10-bit tags enabled. In the kernel we call devices
that talk to things besides host memory "P2PDMA". So the spec is saying
not to enable 10bit tags for devices participating in P2PDMA. The kernel
needs a way to allow users to do that. The kernel parameter only stops
the feature from being enabled for a specific device, and the only
use-case is P2PDMA which is not that common and requires the user to be
aware of their topology. So I really don't think this is that big a problem.

>>
>> This has worked well for disable_acs_redir and was used for
>> resource_alignment before that for quite some time. So save a better
>> suggestion I think this is more than acceptable.
> 
> I don't know about other parameters and their history, but we are not in
> 90s anymore and addition of modules parameters (for the PCI it is kernel
> cmdline arguments) are better to be changed to some configuration tool/sysfs.

The problem was that the ACS bits had to be set before the kernel
enumerated the devices. The IOMMU code simply was not able to support
dynamic adjustments to its groups. I assume changing 10bit tags
dynamically is similarly tricky -- but if it's not then, yes a sysfs
interface in addition to the kernel parameter would be a good idea.

Logan
Leon Romanovsky July 27, 2021, 11:05 a.m. UTC | #7 
On Mon, Jul 26, 2021 at 09:48:57AM -0600, Logan Gunthorpe wrote:
> 
> 
> On 2021-07-25 12:39 a.m., Leon Romanovsky wrote:
> > On Fri, Jul 23, 2021 at 10:20:50AM -0600, Logan Gunthorpe wrote:
> >>
> >>
> >>
> >> On 2021-07-23 5:32 a.m., Leon Romanovsky wrote:
> >>> On Fri, Jul 23, 2021 at 07:06:41PM +0800, Dongdong Liu wrote:
> >>>> PCIe spec 5.0 r1.0 section 2.2.6.2 says that if an Endpoint supports
> >>>> sending Requests to other Endpoints (as opposed to host memory), the
> >>>> Endpoint must not send 10-Bit Tag Requests to another given Endpoint
> >>>> unless an implementation-specific mechanism determines that the Endpoint
> >>>> supports 10-Bit Tag Completer capability. Add "pci=disable_10bit_tag="
> >>>> parameter to disable 10-Bit Tag Requester if the peer device does not
> >>>> support the 10-Bit Tag Completer. This will make P2P traffic safe.
> >>>>
> >>>> Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
> >>>> ---
> >>>>  Documentation/admin-guide/kernel-parameters.txt |  7 ++++
> >>>>  drivers/pci/pci.c                               | 56 +++++++++++++++++++++++++
> >>>>  drivers/pci/pci.h                               |  1 +
> >>>>  drivers/pci/pcie/portdrv_pci.c                  | 13 +++---
> >>>>  drivers/pci/probe.c                             |  9 ++--
> >>>>  5 files changed, 78 insertions(+), 8 deletions(-)
> >>>>
> >>>> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> >>>> index bdb2200..c2c4585 100644
> >>>> --- a/Documentation/admin-guide/kernel-parameters.txt
> >>>> +++ b/Documentation/admin-guide/kernel-parameters.txt
> >>>> @@ -4019,6 +4019,13 @@
> >>>>  				bridges without forcing it upstream. Note:
> >>>>  				this removes isolation between devices and
> >>>>  				may put more devices in an IOMMU group.
> >>>> +		disable_10bit_tag=<pci_dev>[; ...]
> >>>> +				  Specify one or more PCI devices (in the format
> >>>> +				  specified above) separated by semicolons.
> >>>> +				  Disable 10-Bit Tag Requester if the peer
> >>>> +				  device does not support the 10-Bit Tag
> >>>> +				  Completer.This will make P2P traffic safe.
> >>>
> >>> I can't imagine more awkward user experience than such kernel parameter.
> >>>
> >>> As a user, I will need to boot the system, hope for the best that system
> >>> works, write down all PCI device numbers, guess which one doesn't work
> >>> properly, update grub with new command line argument and reboot the
> >>> system. Any HW change and this dance should be repeated.
> >>
> >> There are already two such PCI parameters with this pattern and they are
> >> not that awkward. pci_dev may be specified with either vendor/device IDS
> >> or with a path of BDFs (which protects against renumbering).
> > 
> > Unfortunately, in the real world, BDF is not so stable. It changes with
> > addition of new hardware, BIOS upgrades and even broken servers.
> 
> That's why it supports using a *path* of BDFs which tends not to catch
> the wrong device if the topology changes.
> 
> > Vendor/device IDs doesn't work if you have multiple devices of same
> > vendor in the system.
> 
> Yes, but it's fine for some use cases. That's why there's a range of
> options.

The thing is that you are adding PCI parameter that is applicable to everyone.

We probably see different usage models for this feature. In my world, users
have thousands of servers that runs 24x7, with VMs on top, some of them perform
FW upgrades without stopping anything. The idea that you can reboot such server
any time, simply doesn't exist.

So if I need to enable/disable this feature for one of the VFs, I will be stuck.

> 
> >>
> >> This flag is only useful in P2PDMA traffic, and if the user attempts
> >> such a transfer, it prints a warning (see the next patch) with the exact
> >> parameter that needs to be added to the command line.
> > 
> > Dongdong citied PCI spec and it was very clear - don't enable this
> > feature unless you clearly know that it is safe to enable. This is
> > completely opposite to the proposal here - always enable and disable
> > if something is printed to the dmesg.
> 
> Quoting from patch 4:
> 
> "For platforms where the RC supports 10-Bit Tag Completer capability,
> it is highly recommended for platform firmware or operating software
> that configures PCIe hierarchies to Set the 10-Bit Tag Requester Enable
> bit automatically in Endpoints with 10-Bit Tag Requester capability.
> This enables the important class of 10-Bit Tag capable adapters that
> send Memory Read Requests only to host memory."
> 
> Notice the last sentence. It's saying that devices who only talk to host
> memory should have 10-bit tags enabled. In the kernel we call devices
> that talk to things besides host memory "P2PDMA". So the spec is saying
> not to enable 10bit tags for devices participating in P2PDMA. The kernel
> needs a way to allow users to do that. The kernel parameter only stops
> the feature from being enabled for a specific device, and the only
> use-case is P2PDMA which is not that common and requires the user to be
> aware of their topology. So I really don't think this is that big a problem.

I'm not question the feature and the need of configuration. My concern
is just *how* this feature is configured.

> 
> >>
> >> This has worked well for disable_acs_redir and was used for
> >> resource_alignment before that for quite some time. So save a better
> >> suggestion I think this is more than acceptable.
> > 
> > I don't know about other parameters and their history, but we are not in
> > 90s anymore and addition of modules parameters (for the PCI it is kernel
> > cmdline arguments) are better to be changed to some configuration tool/sysfs.
> 
> The problem was that the ACS bits had to be set before the kernel
> enumerated the devices. The IOMMU code simply was not able to support
> dynamic adjustments to its groups. I assume changing 10bit tags
> dynamically is similarly tricky -- but if it's not then, yes a sysfs
> interface in addition to the kernel parameter would be a good idea.

I think that it is doable with combination of drivers_autoprobe disable
and some sysfs knob to enable/disable this feature before driver bind.

It should be very similar to that we did for the dynamic MSI-X, see
/sys/bus/pci/devices/.../sriov_vf_msix_count

Thanks

> 
> Logan
Dongdong Liu July 27, 2021, 2 p.m. UTC | #8 
On 2021/7/26 23:48, Logan Gunthorpe wrote:
>
>
> On 2021-07-25 12:39 a.m., Leon Romanovsky wrote:
>> On Fri, Jul 23, 2021 at 10:20:50AM -0600, Logan Gunthorpe wrote:
>>>
>>>
>>>
>>> On 2021-07-23 5:32 a.m., Leon Romanovsky wrote:
>>>> On Fri, Jul 23, 2021 at 07:06:41PM +0800, Dongdong Liu wrote:
>>>>> PCIe spec 5.0 r1.0 section 2.2.6.2 says that if an Endpoint supports
>>>>> sending Requests to other Endpoints (as opposed to host memory), the
>>>>> Endpoint must not send 10-Bit Tag Requests to another given Endpoint
>>>>> unless an implementation-specific mechanism determines that the Endpoint
>>>>> supports 10-Bit Tag Completer capability. Add "pci=disable_10bit_tag="
>>>>> parameter to disable 10-Bit Tag Requester if the peer device does not
>>>>> support the 10-Bit Tag Completer. This will make P2P traffic safe.
>>>>>
>>>>> Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
>>>>> ---
>>>>>  Documentation/admin-guide/kernel-parameters.txt |  7 ++++
>>>>>  drivers/pci/pci.c                               | 56 +++++++++++++++++++++++++
>>>>>  drivers/pci/pci.h                               |  1 +
>>>>>  drivers/pci/pcie/portdrv_pci.c                  | 13 +++---
>>>>>  drivers/pci/probe.c                             |  9 ++--
>>>>>  5 files changed, 78 insertions(+), 8 deletions(-)
>>>>>
>>>>> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
>>>>> index bdb2200..c2c4585 100644
>>>>> --- a/Documentation/admin-guide/kernel-parameters.txt
>>>>> +++ b/Documentation/admin-guide/kernel-parameters.txt
>>>>> @@ -4019,6 +4019,13 @@
>>>>>  				bridges without forcing it upstream. Note:
>>>>>  				this removes isolation between devices and
>>>>>  				may put more devices in an IOMMU group.
>>>>> +		disable_10bit_tag=<pci_dev>[; ...]
>>>>> +				  Specify one or more PCI devices (in the format
>>>>> +				  specified above) separated by semicolons.
>>>>> +				  Disable 10-Bit Tag Requester if the peer
>>>>> +				  device does not support the 10-Bit Tag
>>>>> +				  Completer.This will make P2P traffic safe.
>>>>
>>>> I can't imagine more awkward user experience than such kernel parameter.
>>>>
>>>> As a user, I will need to boot the system, hope for the best that system
>>>> works, write down all PCI device numbers, guess which one doesn't work
>>>> properly, update grub with new command line argument and reboot the
>>>> system. Any HW change and this dance should be repeated.
>>>
>>> There are already two such PCI parameters with this pattern and they are
>>> not that awkward. pci_dev may be specified with either vendor/device IDS
>>> or with a path of BDFs (which protects against renumbering).
>>
>> Unfortunately, in the real world, BDF is not so stable. It changes with
>> addition of new hardware, BIOS upgrades and even broken servers.
>
> That's why it supports using a *path* of BDFs which tends not to catch
> the wrong device if the topology changes.
>
>> Vendor/device IDs doesn't work if you have multiple devices of same
>> vendor in the system.
>
> Yes, but it's fine for some use cases. That's why there's a range of
> options.
>
>>>
>>> This flag is only useful in P2PDMA traffic, and if the user attempts
>>> such a transfer, it prints a warning (see the next patch) with the exact
>>> parameter that needs to be added to the command line.
>>
>> Dongdong citied PCI spec and it was very clear - don't enable this
>> feature unless you clearly know that it is safe to enable. This is
>> completely opposite to the proposal here - always enable and disable
>> if something is printed to the dmesg.
>
> Quoting from patch 4:
>
> "For platforms where the RC supports 10-Bit Tag Completer capability,
> it is highly recommended for platform firmware or operating software
> that configures PCIe hierarchies to Set the 10-Bit Tag Requester Enable
> bit automatically in Endpoints with 10-Bit Tag Requester capability.
> This enables the important class of 10-Bit Tag capable adapters that
> send Memory Read Requests only to host memory."
>
> Notice the last sentence. It's saying that devices who only talk to host
> memory should have 10-bit tags enabled. In the kernel we call devices
> that talk to things besides host memory "P2PDMA". So the spec is saying
> not to enable 10bit tags for devices participating in P2PDMA. The kernel
> needs a way to allow users to do that. The kernel parameter only stops
> the feature from being enabled for a specific device, and the only
> use-case is P2PDMA which is not that common and requires the user to be
> aware of their topology. So I really don't think this is that big a problem.
>
>>>
>>> This has worked well for disable_acs_redir and was used for
>>> resource_alignment before that for quite some time. So save a better
>>> suggestion I think this is more than acceptable.
>>
>> I don't know about other parameters and their history, but we are not in
>> 90s anymore and addition of modules parameters (for the PCI it is kernel
>> cmdline arguments) are better to be changed to some configuration tool/sysfs.
>
> The problem was that the ACS bits had to be set before the kernel
> enumerated the devices. The IOMMU code simply was not able to support
> dynamic adjustments to its groups. I assume changing 10bit tags
> dynamically is similarly tricky -- but if it's not then, yes a sysfs
> interface in addition to the kernel parameter would be a good idea.
PCIe spec 5.0 section 7.5.3.16 Device Control 2 Register
10-Bit Tag Requester Enable says that
If software changes the value of this bit while the Function
has outstanding Non-Posted Requests, the result is undefined.

So 10-Bit Tag Requester Enable should be set before probe the device 
driver.

Thanks,
Dongdong
>
> Logan
> .
>
Dongdong Liu July 27, 2021, 2:30 p.m. UTC | #9 
On 2021/7/27 19:05, Leon Romanovsky wrote:
> On Mon, Jul 26, 2021 at 09:48:57AM -0600, Logan Gunthorpe wrote:
>>
>>
>> On 2021-07-25 12:39 a.m., Leon Romanovsky wrote:
>>> On Fri, Jul 23, 2021 at 10:20:50AM -0600, Logan Gunthorpe wrote:
>>>>
>>>>
>>>>
>>>> On 2021-07-23 5:32 a.m., Leon Romanovsky wrote:
>>>>> On Fri, Jul 23, 2021 at 07:06:41PM +0800, Dongdong Liu wrote:
>>>>>> PCIe spec 5.0 r1.0 section 2.2.6.2 says that if an Endpoint supports
>>>>>> sending Requests to other Endpoints (as opposed to host memory), the
>>>>>> Endpoint must not send 10-Bit Tag Requests to another given Endpoint
>>>>>> unless an implementation-specific mechanism determines that the Endpoint
>>>>>> supports 10-Bit Tag Completer capability. Add "pci=disable_10bit_tag="
>>>>>> parameter to disable 10-Bit Tag Requester if the peer device does not
>>>>>> support the 10-Bit Tag Completer. This will make P2P traffic safe.
>>>>>>
>>>>>> Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
>>>>>> ---
>>>>>>  Documentation/admin-guide/kernel-parameters.txt |  7 ++++
>>>>>>  drivers/pci/pci.c                               | 56 +++++++++++++++++++++++++
>>>>>>  drivers/pci/pci.h                               |  1 +
>>>>>>  drivers/pci/pcie/portdrv_pci.c                  | 13 +++---
>>>>>>  drivers/pci/probe.c                             |  9 ++--
>>>>>>  5 files changed, 78 insertions(+), 8 deletions(-)
>>>>>>
>>>>>> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
>>>>>> index bdb2200..c2c4585 100644
>>>>>> --- a/Documentation/admin-guide/kernel-parameters.txt
>>>>>> +++ b/Documentation/admin-guide/kernel-parameters.txt
>>>>>> @@ -4019,6 +4019,13 @@
>>>>>>  				bridges without forcing it upstream. Note:
>>>>>>  				this removes isolation between devices and
>>>>>>  				may put more devices in an IOMMU group.
>>>>>> +		disable_10bit_tag=<pci_dev>[; ...]
>>>>>> +				  Specify one or more PCI devices (in the format
>>>>>> +				  specified above) separated by semicolons.
>>>>>> +				  Disable 10-Bit Tag Requester if the peer
>>>>>> +				  device does not support the 10-Bit Tag
>>>>>> +				  Completer.This will make P2P traffic safe.
>>>>>
>>>>> I can't imagine more awkward user experience than such kernel parameter.
>>>>>
>>>>> As a user, I will need to boot the system, hope for the best that system
>>>>> works, write down all PCI device numbers, guess which one doesn't work
>>>>> properly, update grub with new command line argument and reboot the
>>>>> system. Any HW change and this dance should be repeated.
>>>>
>>>> There are already two such PCI parameters with this pattern and they are
>>>> not that awkward. pci_dev may be specified with either vendor/device IDS
>>>> or with a path of BDFs (which protects against renumbering).
>>>
>>> Unfortunately, in the real world, BDF is not so stable. It changes with
>>> addition of new hardware, BIOS upgrades and even broken servers.
>>
>> That's why it supports using a *path* of BDFs which tends not to catch
>> the wrong device if the topology changes.
>>
>>> Vendor/device IDs doesn't work if you have multiple devices of same
>>> vendor in the system.
>>
>> Yes, but it's fine for some use cases. That's why there's a range of
>> options.
>
> The thing is that you are adding PCI parameter that is applicable to everyone.
>
> We probably see different usage models for this feature. In my world, users
> have thousands of servers that runs 24x7, with VMs on top, some of them perform
> FW upgrades without stopping anything. The idea that you can reboot such server
> any time, simply doesn't exist.
>
> So if I need to enable/disable this feature for one of the VFs, I will be stuck.
>
>>
>>>>
>>>> This flag is only useful in P2PDMA traffic, and if the user attempts
>>>> such a transfer, it prints a warning (see the next patch) with the exact
>>>> parameter that needs to be added to the command line.
>>>
>>> Dongdong citied PCI spec and it was very clear - don't enable this
>>> feature unless you clearly know that it is safe to enable. This is
>>> completely opposite to the proposal here - always enable and disable
>>> if something is printed to the dmesg.
>>
>> Quoting from patch 4:
>>
>> "For platforms where the RC supports 10-Bit Tag Completer capability,
>> it is highly recommended for platform firmware or operating software
>> that configures PCIe hierarchies to Set the 10-Bit Tag Requester Enable
>> bit automatically in Endpoints with 10-Bit Tag Requester capability.
>> This enables the important class of 10-Bit Tag capable adapters that
>> send Memory Read Requests only to host memory."
>>
>> Notice the last sentence. It's saying that devices who only talk to host
>> memory should have 10-bit tags enabled. In the kernel we call devices
>> that talk to things besides host memory "P2PDMA". So the spec is saying
>> not to enable 10bit tags for devices participating in P2PDMA. The kernel
>> needs a way to allow users to do that. The kernel parameter only stops
>> the feature from being enabled for a specific device, and the only
>> use-case is P2PDMA which is not that common and requires the user to be
>> aware of their topology. So I really don't think this is that big a problem.
>
> I'm not question the feature and the need of configuration. My concern
> is just *how* this feature is configured.
>
>>
>>>>
>>>> This has worked well for disable_acs_redir and was used for
>>>> resource_alignment before that for quite some time. So save a better
>>>> suggestion I think this is more than acceptable.
>>>
>>> I don't know about other parameters and their history, but we are not in
>>> 90s anymore and addition of modules parameters (for the PCI it is kernel
>>> cmdline arguments) are better to be changed to some configuration tool/sysfs.
>>
>> The problem was that the ACS bits had to be set before the kernel
>> enumerated the devices. The IOMMU code simply was not able to support
>> dynamic adjustments to its groups. I assume changing 10bit tags
>> dynamically is similarly tricky -- but if it's not then, yes a sysfs
>> interface in addition to the kernel parameter would be a good idea.
>
> I think that it is doable with combination of drivers_autoprobe disable
> and some sysfs knob to enable/disable this feature before driver bind.
>
> It should be very similar to that we did for the dynamic MSI-X, see
> /sys/bus/pci/devices/.../sriov_vf_msix_count

Many thanks for your suggestion.

Seems a sysfs could be work ok,  but need to make sure 10-Bit Tag 
Requester to be set before binding the device driver as
PCIe spec 5.0 section 7.5.3.16 Device Control 2 Register
10-Bit Tag Requester Enable says that
If software changes the value of this bit while the Function
has outstanding Non-Posted Requests, the result is undefined.

Thanks,
Dongdong
>
> Thanks
>
>>
>> Logan
> .
>
Leon Romanovsky July 27, 2021, 3:41 p.m. UTC | #10 
On Tue, Jul 27, 2021 at 10:30:40PM +0800, Dongdong Liu wrote:
> 
> 
> On 2021/7/27 19:05, Leon Romanovsky wrote:
> > On Mon, Jul 26, 2021 at 09:48:57AM -0600, Logan Gunthorpe wrote:
> > > 
> > > 
> > > On 2021-07-25 12:39 a.m., Leon Romanovsky wrote:
> > > > On Fri, Jul 23, 2021 at 10:20:50AM -0600, Logan Gunthorpe wrote:
> > > > > 
> > > > > 
> > > > > 
> > > > > On 2021-07-23 5:32 a.m., Leon Romanovsky wrote:
> > > > > > On Fri, Jul 23, 2021 at 07:06:41PM +0800, Dongdong Liu wrote:
> > > > > > > PCIe spec 5.0 r1.0 section 2.2.6.2 says that if an Endpoint supports
> > > > > > > sending Requests to other Endpoints (as opposed to host memory), the
> > > > > > > Endpoint must not send 10-Bit Tag Requests to another given Endpoint
> > > > > > > unless an implementation-specific mechanism determines that the Endpoint
> > > > > > > supports 10-Bit Tag Completer capability. Add "pci=disable_10bit_tag="
> > > > > > > parameter to disable 10-Bit Tag Requester if the peer device does not
> > > > > > > support the 10-Bit Tag Completer. This will make P2P traffic safe.
> > > > > > > 
> > > > > > > Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
> > > > > > > ---
> > > > > > >  Documentation/admin-guide/kernel-parameters.txt |  7 ++++
> > > > > > >  drivers/pci/pci.c                               | 56 +++++++++++++++++++++++++
> > > > > > >  drivers/pci/pci.h                               |  1 +
> > > > > > >  drivers/pci/pcie/portdrv_pci.c                  | 13 +++---
> > > > > > >  drivers/pci/probe.c                             |  9 ++--
> > > > > > >  5 files changed, 78 insertions(+), 8 deletions(-)
> > > > > > > 
> > > > > > > diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> > > > > > > index bdb2200..c2c4585 100644
> > > > > > > --- a/Documentation/admin-guide/kernel-parameters.txt
> > > > > > > +++ b/Documentation/admin-guide/kernel-parameters.txt
> > > > > > > @@ -4019,6 +4019,13 @@
> > > > > > >  				bridges without forcing it upstream. Note:
> > > > > > >  				this removes isolation between devices and
> > > > > > >  				may put more devices in an IOMMU group.
> > > > > > > +		disable_10bit_tag=<pci_dev>[; ...]
> > > > > > > +				  Specify one or more PCI devices (in the format
> > > > > > > +				  specified above) separated by semicolons.
> > > > > > > +				  Disable 10-Bit Tag Requester if the peer
> > > > > > > +				  device does not support the 10-Bit Tag
> > > > > > > +				  Completer.This will make P2P traffic safe.
> > > > > > 
> > > > > > I can't imagine more awkward user experience than such kernel parameter.
> > > > > > 
> > > > > > As a user, I will need to boot the system, hope for the best that system
> > > > > > works, write down all PCI device numbers, guess which one doesn't work
> > > > > > properly, update grub with new command line argument and reboot the
> > > > > > system. Any HW change and this dance should be repeated.
> > > > > 
> > > > > There are already two such PCI parameters with this pattern and they are
> > > > > not that awkward. pci_dev may be specified with either vendor/device IDS
> > > > > or with a path of BDFs (which protects against renumbering).
> > > > 
> > > > Unfortunately, in the real world, BDF is not so stable. It changes with
> > > > addition of new hardware, BIOS upgrades and even broken servers.
> > > 
> > > That's why it supports using a *path* of BDFs which tends not to catch
> > > the wrong device if the topology changes.
> > > 
> > > > Vendor/device IDs doesn't work if you have multiple devices of same
> > > > vendor in the system.
> > > 
> > > Yes, but it's fine for some use cases. That's why there's a range of
> > > options.
> > 
> > The thing is that you are adding PCI parameter that is applicable to everyone.
> > 
> > We probably see different usage models for this feature. In my world, users
> > have thousands of servers that runs 24x7, with VMs on top, some of them perform
> > FW upgrades without stopping anything. The idea that you can reboot such server
> > any time, simply doesn't exist.
> > 
> > So if I need to enable/disable this feature for one of the VFs, I will be stuck.
> > 
> > > 
> > > > > 
> > > > > This flag is only useful in P2PDMA traffic, and if the user attempts
> > > > > such a transfer, it prints a warning (see the next patch) with the exact
> > > > > parameter that needs to be added to the command line.
> > > > 
> > > > Dongdong citied PCI spec and it was very clear - don't enable this
> > > > feature unless you clearly know that it is safe to enable. This is
> > > > completely opposite to the proposal here - always enable and disable
> > > > if something is printed to the dmesg.
> > > 
> > > Quoting from patch 4:
> > > 
> > > "For platforms where the RC supports 10-Bit Tag Completer capability,
> > > it is highly recommended for platform firmware or operating software
> > > that configures PCIe hierarchies to Set the 10-Bit Tag Requester Enable
> > > bit automatically in Endpoints with 10-Bit Tag Requester capability.
> > > This enables the important class of 10-Bit Tag capable adapters that
> > > send Memory Read Requests only to host memory."
> > > 
> > > Notice the last sentence. It's saying that devices who only talk to host
> > > memory should have 10-bit tags enabled. In the kernel we call devices
> > > that talk to things besides host memory "P2PDMA". So the spec is saying
> > > not to enable 10bit tags for devices participating in P2PDMA. The kernel
> > > needs a way to allow users to do that. The kernel parameter only stops
> > > the feature from being enabled for a specific device, and the only
> > > use-case is P2PDMA which is not that common and requires the user to be
> > > aware of their topology. So I really don't think this is that big a problem.
> > 
> > I'm not question the feature and the need of configuration. My concern
> > is just *how* this feature is configured.
> > 
> > > 
> > > > > 
> > > > > This has worked well for disable_acs_redir and was used for
> > > > > resource_alignment before that for quite some time. So save a better
> > > > > suggestion I think this is more than acceptable.
> > > > 
> > > > I don't know about other parameters and their history, but we are not in
> > > > 90s anymore and addition of modules parameters (for the PCI it is kernel
> > > > cmdline arguments) are better to be changed to some configuration tool/sysfs.
> > > 
> > > The problem was that the ACS bits had to be set before the kernel
> > > enumerated the devices. The IOMMU code simply was not able to support
> > > dynamic adjustments to its groups. I assume changing 10bit tags
> > > dynamically is similarly tricky -- but if it's not then, yes a sysfs
> > > interface in addition to the kernel parameter would be a good idea.
> > 
> > I think that it is doable with combination of drivers_autoprobe disable
> > and some sysfs knob to enable/disable this feature before driver bind.
> > 
> > It should be very similar to that we did for the dynamic MSI-X, see
> > /sys/bus/pci/devices/.../sriov_vf_msix_count
> 
> Many thanks for your suggestion.
> 
> Seems a sysfs could be work ok,  but need to make sure 10-Bit Tag Requester
> to be set before binding the device driver as
> PCIe spec 5.0 section 7.5.3.16 Device Control 2 Register
> 10-Bit Tag Requester Enable says that
> If software changes the value of this bit while the Function
> has outstanding Non-Posted Requests, the result is undefined.

This is where drivers_autoprobe will help.

Thanks


> 
> Thanks,
> Dongdong
> > 
> > Thanks
> > 
> > > 
> > > Logan
> > .
> >
diff mbox series

Patch

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index bdb2200..c2c4585 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -4019,6 +4019,13 @@ 
 				bridges without forcing it upstream. Note:
 				this removes isolation between devices and
 				may put more devices in an IOMMU group.
+		disable_10bit_tag=<pci_dev>[; ...]
+				  Specify one or more PCI devices (in the format
+				  specified above) separated by semicolons.
+				  Disable 10-Bit Tag Requester if the peer
+				  device does not support the 10-Bit Tag
+				  Completer.This will make P2P traffic safe.
+
 		force_floating	[S390] Force usage of floating interrupts.
 		nomio		[S390] Do not use MIO instructions.
 		norid		[S390] ignore the RID field and force use of
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index d14c573..8494e4f 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -6568,6 +6568,59 @@  int pci_bus_find_domain_nr(struct pci_bus *bus, struct device *parent)
 }
 #endif
 
+static const char *disable_10bit_tag_param;
+
+void pci_disable_10bit_tag(struct pci_dev *dev)
+{
+	int ret = 0;
+	const char *p;
+#ifdef CONFIG_PCI_IOV
+	struct pci_sriov *iov;
+#endif
+
+	if (!disable_10bit_tag_param)
+		return;
+
+	p = disable_10bit_tag_param;
+	while (*p) {
+		ret = pci_dev_str_match(dev, p, &p);
+		if (ret < 0) {
+			pr_info_once("PCI: Can't parse disable_10bit_tag parameter: %s\n",
+				     disable_10bit_tag_param);
+
+			break;
+		} else if (ret == 1) {
+			/* Found a match */
+			break;
+		}
+
+		if (*p != ';' && *p != ',') {
+			/* End of param or invalid format */
+			break;
+		}
+		p++;
+	}
+
+	if (ret != 1)
+		return;
+
+#ifdef CONFIG_PCI_IOV
+	if (dev->is_virtfn) {
+		iov = dev->physfn->sriov;
+		iov->ctrl &= ~PCI_SRIOV_CTRL_VF_10BIT_TAG_REQ_EN;
+		pci_write_config_word(dev, iov->pos + PCI_SRIOV_CTRL,
+				      iov->ctrl);
+		pci_info(dev, "disabled PF SRIOV 10-Bit Tag Requester\n");
+		return;
+#endif
+	}
+
+	pcie_capability_clear_word(dev, PCI_EXP_DEVCTL2,
+				   PCI_EXP_DEVCTL2_10BIT_TAG_REQ_EN);
+
+	pci_info(dev, "disabled 10-Bit Tag Requester\n");
+}
+
 /**
  * pci_ext_cfg_avail - can we access extended PCI config space?
  *
@@ -6643,6 +6696,8 @@  static int __init pci_setup(char *str)
 				pci_add_flags(PCI_SCAN_ALL_PCIE_DEVS);
 			} else if (!strncmp(str, "disable_acs_redir=", 18)) {
 				disable_acs_redir_param = str + 18;
+			} else if (!strncmp(str, "disable_10bit_tag=", 18)) {
+				disable_10bit_tag_param = str + 18;
 			} else {
 				pr_err("PCI: Unknown option `%s'\n", str);
 			}
@@ -6667,6 +6722,7 @@  static int __init pci_realloc_setup_params(void)
 	resource_alignment_param = kstrdup(resource_alignment_param,
 					   GFP_KERNEL);
 	disable_acs_redir_param = kstrdup(disable_acs_redir_param, GFP_KERNEL);
+	disable_10bit_tag_param = kstrdup(disable_10bit_tag_param, GFP_KERNEL);
 
 	return 0;
 }
diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
index 93dcdd4..87c8187 100644
--- a/drivers/pci/pci.h
+++ b/drivers/pci/pci.h
@@ -16,6 +16,7 @@  extern bool pci_early_dump;
 
 bool pcie_cap_has_lnkctl(const struct pci_dev *dev);
 bool pcie_cap_has_rtctl(const struct pci_dev *dev);
+void pci_disable_10bit_tag(struct pci_dev *dev);
 
 /* Functions internal to the PCI core code */
 
diff --git a/drivers/pci/pcie/portdrv_pci.c b/drivers/pci/pcie/portdrv_pci.c
index 2382cd2..747728e 100644
--- a/drivers/pci/pcie/portdrv_pci.c
+++ b/drivers/pci/pcie/portdrv_pci.c
@@ -125,15 +125,15 @@  static void pci_configure_rp_10bit_tag(struct pci_dev *dev)
 	bool support = true;
 
 	if (dev->subordinate == NULL)
-		return;
+		goto disable_10bit_tag_req;
 
 	/* If no devices under the root port, no need to enable 10-Bit Tag. */
 	if (list_empty(&dev->subordinate->devices))
-		return;
+		goto disable_10bit_tag_req;
 
 	pci_10bit_tag_comp_support(dev, &support);
 	if (!support)
-		return;
+		goto disable_10bit_tag_req;
 
 	/*
 	 * PCIe spec 5.0r1.0 section 2.2.6.2 implementation note.
@@ -146,14 +146,17 @@  static void pci_configure_rp_10bit_tag(struct pci_dev *dev)
 	 */
 	pci_walk_bus(dev->subordinate, pci_10bit_tag_comp_support, &support);
 	if (!support)
-		return;
+		goto disable_10bit_tag_req;
 
 	if (!(dev->pcie_devcap2 & PCI_EXP_DEVCAP2_10BIT_TAG_REQ))
-		return;
+		goto disable_10bit_tag_req;
 
 	pci_dbg(dev, "enabling 10-Bit Tag Requester\n");
 	pcie_capability_set_word(dev, PCI_EXP_DEVCTL2,
 				 PCI_EXP_DEVCTL2_10BIT_TAG_REQ_EN);
+
+disable_10bit_tag_req:
+	pci_disable_10bit_tag(dev);
 }
 
 /*
diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
index 3da7baa..0b7b053 100644
--- a/drivers/pci/probe.c
+++ b/drivers/pci/probe.c
@@ -2034,11 +2034,11 @@  static void pci_configure_10bit_tags(struct pci_dev *dev)
 	struct pci_dev *bridge;
 
 	if (!(dev->pcie_devcap2 & PCI_EXP_DEVCAP2_10BIT_TAG_COMP))
-		return;
+		goto disable_10bit_tag_req;
 
 	if (pci_pcie_type(dev) == PCI_EXP_TYPE_ROOT_PORT) {
 		dev->ext_10bit_tag = 1;
-		return;
+		goto disable_10bit_tag_req;
 	}
 
 	bridge = pci_upstream_bridge(dev);
@@ -2050,7 +2050,7 @@  static void pci_configure_10bit_tags(struct pci_dev *dev)
 	 * for VF.
 	 */
 	if (dev->is_virtfn)
-		return;
+		goto disable_10bit_tag_req;
 
 	if (pci_pcie_type(dev) == PCI_EXP_TYPE_ENDPOINT &&
 	    dev->ext_10bit_tag == 1 &&
@@ -2059,6 +2059,9 @@  static void pci_configure_10bit_tags(struct pci_dev *dev)
 		pcie_capability_set_word(dev, PCI_EXP_DEVCTL2,
 					PCI_EXP_DEVCTL2_10BIT_TAG_REQ_EN);
 	}
+
+disable_10bit_tag_req:
+	 pci_disable_10bit_tag(dev);
 }
 
 int pci_configure_extended_tags(struct pci_dev *dev, void *ign)