EDAC, skx_edac: Fix for PCI multi-segment system

EDAC, skx_edac: Fix for PCI multi-segment system

[Luck, Tony]

On Thu, Oct 05, 2017 at 12:14:17PM +0200, Borislav Petkov wrote:
> On Thu, Oct 05, 2017 at 06:46:30PM +0900, Masayoshi Mizuma wrote:
> > This patch introduces to check the segment to get_sky_dev()
> > and fix the panic.
> > 
> > Signed-off-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
> > ---
> >  drivers/edac/skx_edac.c |    9 ++++++---
> >  1 files changed, 6 insertions(+), 3 deletions(-)
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/bp/bp.git/commit/?h=for-next&id=88ae80aa609c7db3e3cc98e97e05badbcc6347dc

Looks like your patch is functionally the same as the one
that Boris took from me a few days ago. Funny how an issue
like this sits around for a while, then three people find
it at once (While I was writing the patch as a result of the
first report, a second one came in ... now you found it too!).

Please check that my patch works on your system (it has seen
limited testing in different configurations).

-Tony
---
To unsubscribe from this list: send the line "unsubscribe linux-edac" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

EDAC, skx_edac: Fix for PCI multi-segment system

[Masayoshi Mizuma]

Hi Boris and Tony,

On Thu, 5 Oct 2017 10:26:12 -0700 Luck, Tony wrote:
> On Thu, Oct 05, 2017 at 12:14:17PM +0200, Borislav Petkov wrote:
> > On Thu, Oct 05, 2017 at 06:46:30PM +0900, Masayoshi Mizuma wrote:
> > > This patch introduces to check the segment to get_sky_dev()
> > > and fix the panic.
> > > 
> > > Signed-off-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
> > > ---
> > >  drivers/edac/skx_edac.c |    9 ++++++---
> > >  1 files changed, 6 insertions(+), 3 deletions(-)
> > 
> > https://git.kernel.org/pub/scm/linux/kernel/git/bp/bp.git/commit/?h=for-next&id=88ae80aa609c7db3e3cc98e97e05badbcc6347dc

Thank you for letting me know.

> 
> Looks like your patch is functionally the same as the one
> that Boris took from me a few days ago. Funny how an issue
> like this sits around for a while, then three people find
> it at once (While I was writing the patch as a result of the
> first report, a second one came in ... now you found it too!).
> 
> Please check that my patch works on your system (it has seen
> limited testing in different configurations).

I checked your patch and it works good for me.
Thank you for the patch!

- Masayoshi Mizuma
---
To unsubscribe from this list: send the line "unsubscribe linux-edac" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

EDAC, skx_edac: Fix for PCI multi-segment system

[Borislav Petkov]

On Thu, Oct 05, 2017 at 06:46:30PM +0900, Masayoshi Mizuma wrote:
> Currently, skx_edac is not handled correctly the PCI segment
> except segment #0 because get_sky_dev() is only check the
> bus number of the uncore devices. If the system has milti-segment
> and each bus number of the devices are the same number,
> skx_edac driver initializes the devices only segment #0.
> 
> So, if skx_edac module is loaded to PCI multi-segment system,
> kernel panic occurs because of reference of the uninitialized
> segment #1 or more.
> 
> This patch introduces to check the segment to get_sky_dev()
> and fix the panic.
> 
> Signed-off-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
> ---
>  drivers/edac/skx_edac.c |    9 ++++++---
>  1 files changed, 6 insertions(+), 3 deletions(-)

https://git.kernel.org/pub/scm/linux/kernel/git/bp/bp.git/commit/?h=for-next&id=88ae80aa609c7db3e3cc98e97e05badbcc6347dc

EDAC, skx_edac: Fix for PCI multi-segment system

[Masayoshi Mizuma]

Currently, skx_edac is not handled correctly the PCI segment
except segment #0 because get_sky_dev() is only check the
bus number of the uncore devices. If the system has milti-segment
and each bus number of the devices are the same number,
skx_edac driver initializes the devices only segment #0.

So, if skx_edac module is loaded to PCI multi-segment system,
kernel panic occurs because of reference of the uninitialized
segment #1 or more.

This patch introduces to check the segment to get_sky_dev()
and fix the panic.

Signed-off-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
---
 drivers/edac/skx_edac.c |    9 ++++++---
 1 files changed, 6 insertions(+), 3 deletions(-)

diff --git a/drivers/edac/skx_edac.c b/drivers/edac/skx_edac.c
index 16dea97..7f56044 100644
--- a/drivers/edac/skx_edac.c
+++ b/drivers/edac/skx_edac.c
@@ -68,6 +68,7 @@ struct skx_dev {
 	struct pci_dev	*sad_all;
 	struct pci_dev	*util_all;
 	u32	mcroute;
+	int	domain;
 	struct skx_imc {
 		struct mem_ctl_info *mci;
 		u8	mc;	/* system wide mc# */
@@ -110,12 +111,12 @@ struct decoded_addr {
 	int	bank_group;
 };
 
-static struct skx_dev *get_skx_dev(u8 bus, u8 idx)
+static struct skx_dev *get_skx_dev(int domain, u8 bus, u8 idx)
 {
 	struct skx_dev *d;
 
 	list_for_each_entry(d, &skx_edac_list, list) {
-		if (d->bus[idx] == bus)
+		if ((d->bus[idx] == bus) && (d->domain == domain))
 			return d;
 	}
 
@@ -179,6 +180,7 @@ static int get_all_bus_mappings(void)
 		d->bus[3] =  GET_BITFIELD(reg, 24, 31);
 		edac_dbg(2, "busses: %x, %x, %x, %x\n",
 			 d->bus[0], d->bus[1], d->bus[2], d->bus[3]);
+		d->domain =  pci_domain_nr(pdev->bus);
 		list_add_tail(&d->list, &skx_edac_list);
 		skx_num_sockets++;
 		prev = pdev;
@@ -207,7 +209,8 @@ static int get_all_munits(const struct munit *m)
 			if (i == NUM_IMC)
 				goto fail;
 		}
-		d = get_skx_dev(pdev->bus->number, m->busidx);
+		d = get_skx_dev(pci_domain_nr(pdev->bus), pdev->bus->number,
+				m->busidx);
 		if (!d)
 			goto fail;