| Message ID | 1416310271-4423-1-git-send-email-thomas.petazzoni@free-electrons.com (mailing list archive) |
|---|---|
| State | New, archived |
| Delegated to: | Bjorn Helgaas |
| Headers | show |
On Tue, Nov 18, 2014 at 12:31:11PM +0100, Thomas Petazzoni wrote: > In commit 7ea945f0bb49 ("PCI: Shrink decoding-disabled window while > sizing BARs"), Myron Stowe refactored the code of __pci_read_base() in > order to reduce the amount of time spent with decoding disabled. > > However, contrary to what was said in the commit log, the commit does > introduce some functional change: the pci_size() function that used to > be called *before* the BAR size check is done is now called *after* > the BAR size check is done. > > This causes some failures on certain platforms (namely ARM Marvell EBU > platforms, equipped for example with a PCIe SATA card, or a PCIe USB3 > XHCI controller): > > pci 0000:03:00.0: reg 0x10: can't handle BAR larger than 4GB (size 0xfffffffffff00000) > > This problem didn't exist before this commit, due to pci_size() being > called before doing the PCI BAR size check. Therefore, this commit > fixes the problem by restoring the initial order of the operation, by > calling pci_size() before doing the PCI BAR size check. > > Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> > Fixes: 7ea945f0bb49 ("PCI: Shrink decoding-disabled window while sizing BARs") > --- > Note: this fix has been tested to work correctly for me, and the PCI > messages I get are now identical to the ones I was getting with > 3.18-rc5. However, please review my patch carefully, as I must admit I > do not fully understand all the implications of the change as well as > the code in __pci_read_base(). > > Applies on top of pci/next > > Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> > --- > drivers/pci/probe.c | 16 ++++++++-------- > 1 file changed, 8 insertions(+), 8 deletions(-) I started seeing this same issue on Tegra today and I can confirm that this patch fixes this regression: Tested-by: Thierry Reding <treding@nvidia.com>
Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes: > In commit 7ea945f0bb49 ("PCI: Shrink decoding-disabled window while > sizing BARs"), Myron Stowe refactored the code of __pci_read_base() in > order to reduce the amount of time spent with decoding disabled. > > However, contrary to what was said in the commit log, the commit does > introduce some functional change: the pci_size() function that used to > be called *before* the BAR size check is done is now called *after* > the BAR size check is done. > > This causes some failures on certain platforms (namely ARM Marvell EBU > platforms, equipped for example with a PCIe SATA card, or a PCIe USB3 > XHCI controller): > > pci 0000:03:00.0: reg 0x10: can't handle BAR larger than 4GB (size 0xfffffffffff00000) > > This problem didn't exist before this commit, due to pci_size() being > called before doing the PCI BAR size check. Therefore, this commit > fixes the problem by restoring the initial order of the operation, by > calling pci_size() before doing the PCI BAR size check. > > Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> > Fixes: 7ea945f0bb49 ("PCI: Shrink decoding-disabled window while sizing BARs") I also boot tested this on armada-370-mirabox (multi_v7_defconfig and mvebu_v7_defconfig) where I originally noticed the problem, and it's booting fine again. Tested-by: Kevin Hilman <khilman@linaro.org> and, you forgot something: ;) Reported-by: Kevin Hilman <khilman@linaro.org> Kevin -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, Nov 18, 2014 at 12:31:11PM +0100, Thomas Petazzoni wrote: > In commit 7ea945f0bb49 ("PCI: Shrink decoding-disabled window while > sizing BARs"), Myron Stowe refactored the code of __pci_read_base() in > order to reduce the amount of time spent with decoding disabled. > > However, contrary to what was said in the commit log, the commit does > introduce some functional change: the pci_size() function that used to > be called *before* the BAR size check is done is now called *after* > the BAR size check is done. > > This causes some failures on certain platforms (namely ARM Marvell EBU > platforms, equipped for example with a PCIe SATA card, or a PCIe USB3 > XHCI controller): > > pci 0000:03:00.0: reg 0x10: can't handle BAR larger than 4GB (size 0xfffffffffff00000) > > This problem didn't exist before this commit, due to pci_size() being > called before doing the PCI BAR size check. Therefore, this commit > fixes the problem by restoring the initial order of the operation, by > calling pci_size() before doing the PCI BAR size check. > > Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> > Fixes: 7ea945f0bb49 ("PCI: Shrink decoding-disabled window while sizing BARs") > --- > Note: this fix has been tested to work correctly for me, and the PCI > messages I get are now identical to the ones I was getting with > 3.18-rc5. However, please review my patch carefully, as I must admit I > do not fully understand all the implications of the change as well as > the code in __pci_read_base(). > > Applies on top of pci/next > > Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> I folded essentially this patch into "PCI: Shrink decoding-disabled window while sizing BARs" and re-merged that into my "next" branch. Thanks! Bjorn > --- > drivers/pci/probe.c | 16 ++++++++-------- > 1 file changed, 8 insertions(+), 8 deletions(-) > > diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c > index db16678..6168ca1 100644 > --- a/drivers/pci/probe.c > +++ b/drivers/pci/probe.c > @@ -246,6 +246,14 @@ int __pci_read_base(struct pci_dev *dev, enum pci_bar_type type, > if (!sz64) > goto fail; > > + sz64 = pci_size(l64, sz64, mask64); > + > + if (!sz64) { > + dev_info(&dev->dev, "%sreg 0x%x: invalid BAR (can't size)\n", > + FW_BUG, pos); > + goto fail; > + } > + > if (res->flags & IORESOURCE_MEM_64) { > if ((sizeof(dma_addr_t) < 8 || sizeof(resource_size_t) < 8) && > sz64 > 0x100000000ULL) { > @@ -268,14 +276,6 @@ int __pci_read_base(struct pci_dev *dev, enum pci_bar_type type, > } > } > > - sz64 = pci_size(l64, sz64, mask64); > - > - if (!sz64) { > - dev_info(&dev->dev, "%sreg 0x%x: invalid BAR (can't size)\n", > - FW_BUG, pos); > - goto fail; > - } > - > region.start = l64; > region.end = l64 + sz64; > > -- > 2.1.0 > -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c index db16678..6168ca1 100644 --- a/drivers/pci/probe.c +++ b/drivers/pci/probe.c @@ -246,6 +246,14 @@ int __pci_read_base(struct pci_dev *dev, enum pci_bar_type type, if (!sz64) goto fail; + sz64 = pci_size(l64, sz64, mask64); + + if (!sz64) { + dev_info(&dev->dev, "%sreg 0x%x: invalid BAR (can't size)\n", + FW_BUG, pos); + goto fail; + } + if (res->flags & IORESOURCE_MEM_64) { if ((sizeof(dma_addr_t) < 8 || sizeof(resource_size_t) < 8) && sz64 > 0x100000000ULL) { @@ -268,14 +276,6 @@ int __pci_read_base(struct pci_dev *dev, enum pci_bar_type type, } } - sz64 = pci_size(l64, sz64, mask64); - - if (!sz64) { - dev_info(&dev->dev, "%sreg 0x%x: invalid BAR (can't size)\n", - FW_BUG, pos); - goto fail; - } - region.start = l64; region.end = l64 + sz64;
In commit 7ea945f0bb49 ("PCI: Shrink decoding-disabled window while sizing BARs"), Myron Stowe refactored the code of __pci_read_base() in order to reduce the amount of time spent with decoding disabled. However, contrary to what was said in the commit log, the commit does introduce some functional change: the pci_size() function that used to be called *before* the BAR size check is done is now called *after* the BAR size check is done. This causes some failures on certain platforms (namely ARM Marvell EBU platforms, equipped for example with a PCIe SATA card, or a PCIe USB3 XHCI controller): pci 0000:03:00.0: reg 0x10: can't handle BAR larger than 4GB (size 0xfffffffffff00000) This problem didn't exist before this commit, due to pci_size() being called before doing the PCI BAR size check. Therefore, this commit fixes the problem by restoring the initial order of the operation, by calling pci_size() before doing the PCI BAR size check. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Fixes: 7ea945f0bb49 ("PCI: Shrink decoding-disabled window while sizing BARs") --- Note: this fix has been tested to work correctly for me, and the PCI messages I get are now identical to the ones I was getting with 3.18-rc5. However, please review my patch carefully, as I must admit I do not fully understand all the implications of the change as well as the code in __pci_read_base(). Applies on top of pci/next Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> --- drivers/pci/probe.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)