Build breakage for pci-next since v4.6-rc5

PATCH New Lukas Wunner 4 participants 5 messages May 13, 2016 – May 17, 2016

Subsystems (2)

PCI IOMMU

lore

Share & tools

lore

Lukas Wunner

author

2016-05-13 11:53 UTC #0 lore raw

Hi Jacek, Hi Jörg,

in case this isn't on your radar yet, Bjorn Helgaas' next branch currently no longer builds when merged with Linus' master branch because this commit introduced with v4.6-rc5:

commit e3156048346c28c695f5cf9db67a8cf88c90f947
Author: Joerg Roedel <[email protected]>
Date: Fri Apr 8 15:12:24 2016 +0200
iommu/amd: Fix checking of pci dma aliases

adds an occurrence of PCI_DEV_FLAGS_DMA_ALIAS_DEVFN but this commit on Bjorn's next branch:

commit 338c3149a221527e202ee26b1e35f76c965bb6c0
Author: Jacek Lawrynowicz <[email protected]>
Date: Thu Mar 3 15:38:02 2016 +0100
PCI: Add support for multiple DMA aliases

removed all other occurrences of this macro. A rebase of the latter commit is required to fix the build breakage. I was made aware of this yesterday by Intel's 0-DAY kernel test infrastructure (see below).

Thanks,

Lukas

----- Forwarded message from kbuild test robot <[email protected]> -----

Date: Thu, 12 May 2016 21:38:46 +0800
From: kbuild test robot <[email protected]>
To: Bjorn Helgaas <[email protected]>
Cc: [email protected], Lukas Wunner <[email protected]>
Subject: [l1k:thunderbolt_runpm_v2 1/14] drivers/iommu/amd_iommu.c:266:22:
error: 'PCI_DEV_FLAGS_DMA_ALIAS_DEVFN' undeclared
User-Agent: Mutt/1.5.23 (2014-03-12)

tree:   https://github.com/l1k/linux thunderbolt_runpm_v2
head:   308f7c892de104a57f50dda4761c10c094f5e80d
commit: 376a0f25bf07bdd90c58985f210c95a59d21fe1b [1/14] Merge remote-tracking branches 'helgaas/next', 'helgaas/for-linus' and 'helgaas/pci/pm' into HEAD
config: x86_64-rhel (attached as .config)
compiler: gcc-4.9 (Debian 4.9.3-14) 4.9.3
reproduce:
        git checkout 376a0f25bf07bdd90c58985f210c95a59d21fe1b
        # save the attached .config to linux build tree
        make ARCH=x86_64

Note: the l1k/thunderbolt_runpm_v2 HEAD 308f7c892de104a57f50dda4761c10c094f5e80d builds fine.
It only hurts bisectibility.

All errors (new ones prefixed by >>):

drivers/iommu/amd_iommu.c: In function 'get_alias':

Quoted 1 line

Quoted 1 line
drivers/iommu/amd_iommu.c:266:22: error: 'PCI_DEV_FLAGS_DMA_ALIAS_DEVFN' undeclared (first use in this function)

      pdev->dev_flags |= PCI_DEV_FLAGS_DMA_ALIAS_DEVFN;
                         ^

drivers/iommu/amd_iommu.c:266:22: note: each undeclared identifier is reported only once for each function it appears in

Quoted 1 line

Quoted 1 line
drivers/iommu/amd_iommu.c:267:7: error: 'struct pci_dev' has no member named 'dma_alias_devfn'

      pdev->dma_alias_devfn = ivrs_alias & 0xff;
          ^

vim +/PCI_DEV_FLAGS_DMA_ALIAS_DEVFN +266 drivers/iommu/amd_iommu.c

e3156048 Joerg Roedel 2016-04-08  260  	/*
e3156048 Joerg Roedel 2016-04-08  261  	 * If we don't have a PCI DMA alias and the IVRS alias is on the same
e3156048 Joerg Roedel 2016-04-08  262  	 * bus, then the IVRS table may know about a quirk that we don't.
e3156048 Joerg Roedel 2016-04-08  263  	 */
e3156048 Joerg Roedel 2016-04-08  264  	if (pci_alias == devid &&
e3156048 Joerg Roedel 2016-04-08  265  	    PCI_BUS_NUM(ivrs_alias) == pdev->bus->number) {
e3156048 Joerg Roedel 2016-04-08 @266  		pdev->dev_flags |= PCI_DEV_FLAGS_DMA_ALIAS_DEVFN;
e3156048 Joerg Roedel 2016-04-08 @267  		pdev->dma_alias_devfn = ivrs_alias & 0xff;
e3156048 Joerg Roedel 2016-04-08  268  		pr_info("AMD-Vi: Added PCI DMA alias %02x.%d for %s\n",
e3156048 Joerg Roedel 2016-04-08  269  			PCI_SLOT(ivrs_alias), PCI_FUNC(ivrs_alias),
e3156048 Joerg Roedel 2016-04-08  270  			dev_name(dev));

:::::: The code at line 266 was first introduced by commit
:::::: e3156048346c28c695f5cf9db67a8cf88c90f947 iommu/amd: Fix checking of pci dma aliases

:::::: TO: Joerg Roedel <[email protected]>
:::::: CC: Joerg Roedel <[email protected]>

---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation

----- End forwarded message -----

Lawrynowicz, Jacek

2016-05-13 12:25 UTC #1 lore raw

Lukas, thanks for the heads up.

The actual commit that needs rebasing is:
commit f0af9593372abfde34460aa1250e670cc535a7d8
Author: Bjorn Helgaas <[email protected]>
Date: Wed Feb 24 13:43:45 2016 -0600
PCI: Add pci_add_dma_alias() to abstract implementation

It changes all direct usages of PCI_DEV_FLAGS_DMA_ALIAS_DEVFN to a pci_add_dma_alias function call.

Bjorn, you want me to post the rebased version?

Regards,
Jacek

-----Original Message-----
From: Lukas Wunner [mailto:[email protected]]
Sent: Friday, May 13, 2016 1:57 PM
To: Joerg Roedel <[email protected]>; Lawrynowicz, Jacek <[email protected]>
Cc: Bjorn Helgaas <[email protected]>; Alex Williamson <[email protected]>; [email protected]
Subject: Build breakage for pci-next since v4.6-rc5

Hi Jacek, Hi Jörg,

in case this isn't on your radar yet, Bjorn Helgaas' next branch currently no longer builds when merged with Linus' master branch because this commit introduced with v4.6-rc5:

commit e3156048346c28c695f5cf9db67a8cf88c90f947
Author: Joerg Roedel <[email protected]>
Date: Fri Apr 8 15:12:24 2016 +0200
iommu/amd: Fix checking of pci dma aliases

adds an occurrence of PCI_DEV_FLAGS_DMA_ALIAS_DEVFN but this commit on Bjorn's next branch:

commit 338c3149a221527e202ee26b1e35f76c965bb6c0
Author: Jacek Lawrynowicz <[email protected]>
Date: Thu Mar 3 15:38:02 2016 +0100
PCI: Add support for multiple DMA aliases

Thanks,

Lukas

----- Forwarded message from kbuild test robot <[email protected]> -----

tree:   https://github.com/l1k/linux thunderbolt_runpm_v2
head:   308f7c892de104a57f50dda4761c10c094f5e80d
commit: 376a0f25bf07bdd90c58985f210c95a59d21fe1b [1/14] Merge remote-tracking

branches 'helgaas/next', 'helgaas/for-linus' and 'helgaas/pci/pm' into HEAD
config: x86_64-rhel (attached as .config)
compiler: gcc-4.9 (Debian 4.9.3-14) 4.9.3
reproduce:

        git checkout 376a0f25bf07bdd90c58985f210c95a59d21fe1b
        # save the attached .config to linux build tree
        make ARCH=x86_64

Note: the l1k/thunderbolt_runpm_v2 HEAD 308f7c892de104a57f50dda4761c10c094f5e80d builds fine.
It only hurts bisectibility.

All errors (new ones prefixed by >>):

drivers/iommu/amd_iommu.c: In function 'get_alias':

Quoted 1 line

Quoted 1 line
drivers/iommu/amd_iommu.c:266:22: error: 'PCI_DEV_FLAGS_DMA_ALIAS_DEVFN'

undeclared (first use in this function)

      pdev->dev_flags |= PCI_DEV_FLAGS_DMA_ALIAS_DEVFN;
                         ^

drivers/iommu/amd_iommu.c:266:22: note: each undeclared identifier is
reported only once for each function it appears in

Quoted 1 line

Quoted 1 line
drivers/iommu/amd_iommu.c:267:7: error: 'struct pci_dev' has no member named

'dma_alias_devfn'

      pdev->dma_alias_devfn = ivrs_alias & 0xff;
          ^

vim +/PCI_DEV_FLAGS_DMA_ALIAS_DEVFN +266 drivers/iommu/amd_iommu.c

e3156048 Joerg Roedel 2016-04-08  260  	/*
e3156048 Joerg Roedel 2016-04-08  261  	 * If we don't have a PCI DMA alias and
the IVRS alias is on the same
e3156048 Joerg Roedel 2016-04-08  262  	 * bus, then the IVRS table may know
about a quirk that we don't.
e3156048 Joerg Roedel 2016-04-08  263  	 */
e3156048 Joerg Roedel 2016-04-08  264  	if (pci_alias == devid &&
e3156048 Joerg Roedel 2016-04-08  265  	    PCI_BUS_NUM(ivrs_alias) ==

pdev->bus->number) {
e3156048 Joerg Roedel 2016-04-08 @266 pdev->dev_flags |=
PCI_DEV_FLAGS_DMA_ALIAS_DEVFN;
e3156048 Joerg Roedel 2016-04-08 @267 pdev->dma_alias_devfn =
ivrs_alias & 0xff;
e3156048 Joerg Roedel 2016-04-08 268 pr_info("AMD-Vi: Added PCI DMA
alias %02x.%d for %s\n",
e3156048 Joerg Roedel 2016-04-08 269 PCI_SLOT(ivrs_alias),
PCI_FUNC(ivrs_alias),
e3156048 Joerg Roedel 2016-04-08 270 dev_name(dev));

:::::: The code at line 266 was first introduced by commit
:::::: e3156048346c28c695f5cf9db67a8cf88c90f947 iommu/amd: Fix checking of pci dma aliases

:::::: TO: Joerg Roedel <[email protected]>
:::::: CC: Joerg Roedel <[email protected]>

---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation

----- End forwarded message -----

Bjorn Helgaas

2016-05-16 20:10 UTC #2 lore raw

Quoted 8 lines · Jacek

Lukas, thanks for the heads up.
The actual commit that needs rebasing is:
commit f0af9593372abfde34460aa1250e670cc535a7d8
Author: Bjorn Helgaas <[email protected]>
Date: Wed Feb 24 13:43:45 2016 -0600
PCI: Add pci_add_dma_alias() to abstract implementation

Quoted 4 lines

It changes all direct usages of PCI_DEV_FLAGS_DMA_ALIAS_DEVFN to a pci_add_dma_alias function call.
Bjorn, you want me to post the rebased version?

I don't want to rebase my "next" branch, but if you could post the incremental patch to fix up drivers/iommu/amd_iommu.c, I'll ask Linus to include that when he pulls my branch.

Quoted 111 lines

-----Original Message-----
From: Lukas Wunner [mailto:[email protected]]
Sent: Friday, May 13, 2016 1:57 PM
To: Joerg Roedel <[email protected]>; Lawrynowicz, Jacek <[email protected]>
Cc: Bjorn Helgaas <[email protected]>; Alex Williamson <[email protected]>; [email protected]
Subject: Build breakage for pci-next since v4.6-rc5
Hi Jacek, Hi Jörg,
in case this isn't on your radar yet, Bjorn Helgaas' next branch currently no longer builds when merged with Linus' master branch because this commit introduced with v4.6-rc5:
commit e3156048346c28c695f5cf9db67a8cf88c90f947
Author: Joerg Roedel <[email protected]>
Date: Fri Apr 8 15:12:24 2016 +0200
iommu/amd: Fix checking of pci dma aliases
adds an occurrence of PCI_DEV_FLAGS_DMA_ALIAS_DEVFN but this commit on Bjorn's next branch:
commit 338c3149a221527e202ee26b1e35f76c965bb6c0
Author: Jacek Lawrynowicz <[email protected]>
Date: Thu Mar 3 15:38:02 2016 +0100
PCI: Add support for multiple DMA aliases
removed all other occurrences of this macro. A rebase of the latter commit is required to fix the build breakage. I was made aware of this yesterday by Intel's 0-DAY kernel test infrastructure (see below).
Thanks,
Lukas
----- Forwarded message from kbuild test robot <[email protected]> -----
Date: Thu, 12 May 2016 21:38:46 +0800
From: kbuild test robot <[email protected]>
To: Bjorn Helgaas <[email protected]>
Cc: [email protected], Lukas Wunner <[email protected]>
Subject: [l1k:thunderbolt_runpm_v2 1/14] drivers/iommu/amd_iommu.c:266:22:
error: 'PCI_DEV_FLAGS_DMA_ALIAS_DEVFN' undeclared
User-Agent: Mutt/1.5.23 (2014-03-12)
tree:   https://github.com/l1k/linux thunderbolt_runpm_v2
head:   308f7c892de104a57f50dda4761c10c094f5e80d
commit: 376a0f25bf07bdd90c58985f210c95a59d21fe1b [1/14] Merge remote-tracking
branches 'helgaas/next', 'helgaas/for-linus' and 'helgaas/pci/pm' into HEAD
config: x86_64-rhel (attached as .config)
compiler: gcc-4.9 (Debian 4.9.3-14) 4.9.3
reproduce:
        git checkout 376a0f25bf07bdd90c58985f210c95a59d21fe1b
        # save the attached .config to linux build tree
        make ARCH=x86_64 
Note: the l1k/thunderbolt_runpm_v2 HEAD 308f7c892de104a57f50dda4761c10c094f5e80d builds fine.
It only hurts bisectibility.
All errors (new ones prefixed by >>):
drivers/iommu/amd_iommu.c: In function 'get_alias':
Quoted 1 line
Quoted 1 line
drivers/iommu/amd_iommu.c:266:22: error: 'PCI_DEV_FLAGS_DMA_ALIAS_DEVFN'
undeclared (first use in this function)
      pdev->dev_flags |= PCI_DEV_FLAGS_DMA_ALIAS_DEVFN;
                         ^
drivers/iommu/amd_iommu.c:266:22: note: each undeclared identifier is
reported only once for each function it appears in
Quoted 1 line
Quoted 1 line
drivers/iommu/amd_iommu.c:267:7: error: 'struct pci_dev' has no member named
'dma_alias_devfn'
      pdev->dma_alias_devfn = ivrs_alias & 0xff;
          ^
vim +/PCI_DEV_FLAGS_DMA_ALIAS_DEVFN +266 drivers/iommu/amd_iommu.c
e3156048 Joerg Roedel 2016-04-08  260  	/*
e3156048 Joerg Roedel 2016-04-08  261  	 * If we don't have a PCI DMA alias and
the IVRS alias is on the same
e3156048 Joerg Roedel 2016-04-08  262  	 * bus, then the IVRS table may know
about a quirk that we don't.
e3156048 Joerg Roedel 2016-04-08  263  	 */
e3156048 Joerg Roedel 2016-04-08  264  	if (pci_alias == devid &&
e3156048 Joerg Roedel 2016-04-08  265  	    PCI_BUS_NUM(ivrs_alias) ==
pdev->bus->number) {
e3156048 Joerg Roedel 2016-04-08 @266 pdev->dev_flags |=
PCI_DEV_FLAGS_DMA_ALIAS_DEVFN;
e3156048 Joerg Roedel 2016-04-08 @267 pdev->dma_alias_devfn =
ivrs_alias & 0xff;
e3156048 Joerg Roedel 2016-04-08 268 pr_info("AMD-Vi: Added PCI DMA
alias %02x.%d for %s\n",
e3156048 Joerg Roedel 2016-04-08 269 PCI_SLOT(ivrs_alias),
PCI_FUNC(ivrs_alias),
e3156048 Joerg Roedel 2016-04-08 270 dev_name(dev));
:::::: The code at line 266 was first introduced by commit
:::::: e3156048346c28c695f5cf9db67a8cf88c90f947 iommu/amd: Fix checking of pci dma aliases
:::::: TO: Joerg Roedel <[email protected]>
:::::: CC: Joerg Roedel <[email protected]>
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
----- End forwarded message -----

Jacek Lawrynowicz

2016-05-17 09:55 UTC #3 lore raw

PATCH 1/1 PCI: Fix merge conflict in amd_iommu.c

Signed-off-by Jacek Lawrynowicz <[email protected]>

---
 drivers/iommu/amd_iommu.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

drivers/iommu/amd_iommu.c +1 -2

index 5efadad..3839fd2 100644
@@ -263,8 +263,7 @@ static u16 get_alias(struct device *dev)
 	 */
 	if (pci_alias == devid &&
 	    PCI_BUS_NUM(ivrs_alias) == pdev->bus->number) {
-		pdev->dev_flags |= PCI_DEV_FLAGS_DMA_ALIAS_DEVFN;
-		pdev->dma_alias_devfn = ivrs_alias & 0xff;
+		pci_add_dma_alias(pdev, ivrs_alias & 0xff);
 		pr_info("AMD-Vi: Added PCI DMA alias %02x.%d for %s\n",
 			PCI_SLOT(ivrs_alias), PCI_FUNC(ivrs_alias),
 			dev_name(dev));

Signature · 1 line

1.8.3.1

Joerg Roedel

2016-05-17 10:17 UTC #4 lore raw

In reply to Bjorn Helgaas

Hi Bjorn,

Quoted 3 lines · Bjorn Helgaas

I don't want to rebase my "next" branch, but if you could post the incremental patch to fix up drivers/iommu/amd_iommu.c, I'll ask Linus to include that when he pulls my branch.

Here is the proper fix:

Quoted 1 line

From 4b19e8719c47fc8fef7c14410db606d16c9c76e3 Mon Sep 17 00:00:00 2001

From: Joerg Roedel <[email protected]>
Date: Tue, 17 May 2016 12:13:57 +0200
Subject: [PATCH] iommu/amd: Use pci_dma_add_alias function

The pci_dma_add_alias() function is the new interface and replaces the old direct-access of 'struct pci_dev'. Convert the last use of the old interface to the new one in the AMD IOMMU driver.

Signed-off-by Joerg Roedel <[email protected]>

---
 drivers/iommu/amd_iommu.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

drivers/iommu/amd_iommu.c +1 -2

index 5efadad..3839fd2 100644
@@ -263,8 +263,7 @@ static u16 get_alias(struct device *dev)
 	 */
 	if (pci_alias == devid &&
 	    PCI_BUS_NUM(ivrs_alias) == pdev->bus->number) {
-		pdev->dev_flags |= PCI_DEV_FLAGS_DMA_ALIAS_DEVFN;
-		pdev->dma_alias_devfn = ivrs_alias & 0xff;
+		pci_add_dma_alias(pdev, ivrs_alias & 0xff);
 		pr_info("AMD-Vi: Added PCI DMA alias %02x.%d for %s\n",
 			PCI_SLOT(ivrs_alias), PCI_FUNC(ivrs_alias),
 			dev_name(dev));

Signature · 1 line

2.6.6