lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20211019224406.GA2414443@bhelgaas>
Date:   Tue, 19 Oct 2021 17:44:06 -0500
From:   Bjorn Helgaas <helgaas@...nel.org>
To:     Naveen Naidu <naveennaidu479@...il.com>
Cc:     bhelgaas@...gle.com, tsbogend@...ha.franken.de, ruscur@...sell.cc,
        oohall@...il.com, linux-pci@...r.kernel.org,
        linuxppc-dev@...ts.ozlabs.org, linux-mips@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        linux-kernel-mentees@...ts.linuxfoundation.org
Subject: Re: [PATCH 1/6] PCI/AER: Enable COR/UNCOR error reporting in
 set_device_error_reporting()

On Mon, Oct 04, 2021 at 11:29:27PM +0530, Naveen Naidu wrote:
> The (PCIe r5.0, sec 7.6.4.3, Table 7-101) and  (PCIe r5.0, sec 7.8.4.6,
> Table 7-104) 

s/7.6.4.3/7.8.4.3/

Cite it like this:

  Per PCIe r5.0, sec 7.8.4.3 and sec 7.8.4.6, the default values ...

> states that the default values for the Uncorrectable Error
> Mask and Correctable Error Mask should be 0b. But the current code does
> not set the default value of these registers when the PCIe bus loads the
> AER service driver.

The defaults specified here are for hardware designers -- this is what
the registers must contain after power-up or reset.  This section of
the spec isn't telling us what the OS is required to write.

If we want to clear these masks, I think we have to:

  1) Analyze every other place that writes the masks to make sure we
  don't break any of them.  There aren't very many, and most of them
  are in drivers, which would be after the aer_probe() path.  There
  might be a conflict with program_hpx_type2(), though.

  2) Make it dependent on pcie_aer_is_native().  Ownership of the AER
  capability can be retained by the platform, in which case the OS
  shouldn't touch it.

> Enable reporting of all correctable and uncorrectable errors during
> aer_probe()
> 
> Signed-off-by: Naveen Naidu <naveennaidu479@...il.com>
> ---
>  drivers/pci/pcie/aer.c | 13 ++++++++++++-
>  1 file changed, 12 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/pci/pcie/aer.c b/drivers/pci/pcie/aer.c
> index 9784fdcf3006..88c4ca6098fb 100644
> --- a/drivers/pci/pcie/aer.c
> +++ b/drivers/pci/pcie/aer.c
> @@ -1212,6 +1212,7 @@ static int set_device_error_reporting(struct pci_dev *dev, void *data)
>  {
>  	bool enable = *((bool *)data);
>  	int type = pci_pcie_type(dev);
> +	int aer = dev->aer_cap;
>  
>  	if ((type == PCI_EXP_TYPE_ROOT_PORT) ||
>  	    (type == PCI_EXP_TYPE_RC_EC) ||
> @@ -1223,8 +1224,18 @@ static int set_device_error_reporting(struct pci_dev *dev, void *data)
>  			pci_disable_pcie_error_reporting(dev);
>  	}
>  
> -	if (enable)
> +	if (enable) {
> +
> +		/* Enable reporting of all uncorrectable errors */
> +		/* Uncorrectable Error Mask - turned on bits disable errors */
> +		pci_write_config_dword(dev, aer + PCI_ERR_UNCOR_MASK, 0);
> +
> +		/* Enable reporting of all correctable errors */
> +		/* Correctable Error Mask - turned on bits disable errors */
> +		pci_write_config_dword(dev, aer + PCI_ERR_COR_MASK, 0);
> +
>  		pcie_set_ecrc_checking(dev);
> +	}
>  
>  	return 0;
>  }
> -- 
> 2.25.1
> 
> _______________________________________________
> Linux-kernel-mentees mailing list
> Linux-kernel-mentees@...ts.linuxfoundation.org
> https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ