| Message ID | 1409911806-10519-4-git-send-email-wangyijing@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Am Freitag, den 05.09.2014, 18:09 +0800 schrieb Yijing Wang: > Currently, pcie-designware, pcie-rcar, pci-tegra drivers > use irq chip_data to save the msi_chip pointer. They > already call irq_set_chip_data() in their own MSI irq map > functions. So irq_set_chip_data() in arch_setup_msi_irq() > is useless. > > Signed-off-by: Yijing Wang <wangyijing@huawei.com> > --- > drivers/pci/msi.c | 2 -- > 1 files changed, 0 insertions(+), 2 deletions(-) > > diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c > index f6cb317..d547f7f 100644 > --- a/drivers/pci/msi.c > +++ b/drivers/pci/msi.c > @@ -41,8 +41,6 @@ int __weak arch_setup_msi_irq(struct pci_dev *dev, struct msi_desc *desc) > if (err < 0) > return err; > > - irq_set_chip_data(desc->irq, chip); > - > return 0; > } > arch_teardown_msi_irq() expects to find the msi_chip in the irq chip_data field. As this means drivers don't have any reasonable other possibility to stuff things into this field, I think it would make sense to do the cleanup the other way around: keep the irq_set_chip_data arch_setup_msi_irq() and rip it out of the individual drivers. Regards, Lucas
On 2014/9/15 22:00, Lucas Stach wrote: > Am Freitag, den 05.09.2014, 18:09 +0800 schrieb Yijing Wang: >> Currently, pcie-designware, pcie-rcar, pci-tegra drivers >> use irq chip_data to save the msi_chip pointer. They >> already call irq_set_chip_data() in their own MSI irq map >> functions. So irq_set_chip_data() in arch_setup_msi_irq() >> is useless. >> >> Signed-off-by: Yijing Wang <wangyijing@huawei.com> >> --- >> drivers/pci/msi.c | 2 -- >> 1 files changed, 0 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c >> index f6cb317..d547f7f 100644 >> --- a/drivers/pci/msi.c >> +++ b/drivers/pci/msi.c >> @@ -41,8 +41,6 @@ int __weak arch_setup_msi_irq(struct pci_dev *dev, struct msi_desc *desc) >> if (err < 0) >> return err; >> >> - irq_set_chip_data(desc->irq, chip); >> - >> return 0; >> } >> > > arch_teardown_msi_irq() expects to find the msi_chip in the irq > chip_data field. As this means drivers don't have any reasonable other > possibility to stuff things into this field, I think it would make sense > to do the cleanup the other way around: keep the irq_set_chip_data > arch_setup_msi_irq() and rip it out of the individual drivers. Hi Lucas, thanks for your review and comments! irq_set_chip_data() should not be placed in MSI core functions, because other arch like x86, use irq_data->chip_data to stores irq_cfg. So how to set the chip_data is arch dependent. And this series is mainly to use MSI chip framework in all platforms. Currently, only ARM platform MSI drivers use the chip_data to store msi_chip, and the drivers call irq_set_chip_data() in their driver already. So I thought we should clean up it in MSI core code. Thanks! Yijing. > > Regards, > Lucas >
Am Dienstag, den 16.09.2014, 09:30 +0800 schrieb Yijing Wang: > On 2014/9/15 22:00, Lucas Stach wrote: > > Am Freitag, den 05.09.2014, 18:09 +0800 schrieb Yijing Wang: > >> Currently, pcie-designware, pcie-rcar, pci-tegra drivers > >> use irq chip_data to save the msi_chip pointer. They > >> already call irq_set_chip_data() in their own MSI irq map > >> functions. So irq_set_chip_data() in arch_setup_msi_irq() > >> is useless. > >> > >> Signed-off-by: Yijing Wang <wangyijing@huawei.com> > >> --- > >> drivers/pci/msi.c | 2 -- > >> 1 files changed, 0 insertions(+), 2 deletions(-) > >> > >> diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c > >> index f6cb317..d547f7f 100644 > >> --- a/drivers/pci/msi.c > >> +++ b/drivers/pci/msi.c > >> @@ -41,8 +41,6 @@ int __weak arch_setup_msi_irq(struct pci_dev *dev, struct msi_desc *desc) > >> if (err < 0) > >> return err; > >> > >> - irq_set_chip_data(desc->irq, chip); > >> - > >> return 0; > >> } > >> > > > > arch_teardown_msi_irq() expects to find the msi_chip in the irq > > chip_data field. As this means drivers don't have any reasonable other > > possibility to stuff things into this field, I think it would make sense > > to do the cleanup the other way around: keep the irq_set_chip_data > > arch_setup_msi_irq() and rip it out of the individual drivers. > > Hi Lucas, thanks for your review and comments! > irq_set_chip_data() should not be placed in MSI core functions, because other arch like x86, > use irq_data->chip_data to stores irq_cfg. So how to set the chip_data is arch dependent. > And this series is mainly to use MSI chip framework in all platforms. > Currently, only ARM platform MSI drivers use the chip_data to store msi_chip, and the drivers call > irq_set_chip_data() in their driver already. So I thought we should clean up it in MSI core code. > Okay I see your point, so the cleanup done this way is okay. But then this still introduces a problem: arch_teardown_msi_irq() expects to find the msi_chip in the chip_data field, which is okay for all ARM PCI host drivers, but not for other arch MSI chips. You fix this by completely removing arch_teardown_msi_irq() at the end of the series, but this still has the potential to introduce issues for other arches than ARM within the series. So this patch should include a change to replace the line struct msi_chip *chip = irq_get_chip_data(irq); with something that doesn't rely on the msi_chip being in the irq chip_data field. Regards, Lucas
>>> arch_teardown_msi_irq() expects to find the msi_chip in the irq >>> chip_data field. As this means drivers don't have any reasonable other >>> possibility to stuff things into this field, I think it would make sense >>> to do the cleanup the other way around: keep the irq_set_chip_data >>> arch_setup_msi_irq() and rip it out of the individual drivers. >> >> Hi Lucas, thanks for your review and comments! >> irq_set_chip_data() should not be placed in MSI core functions, because other arch like x86, >> use irq_data->chip_data to stores irq_cfg. So how to set the chip_data is arch dependent. >> And this series is mainly to use MSI chip framework in all platforms. >> Currently, only ARM platform MSI drivers use the chip_data to store msi_chip, and the drivers call >> irq_set_chip_data() in their driver already. So I thought we should clean up it in MSI core code. >> > Okay I see your point, so the cleanup done this way is okay. > > But then this still introduces a problem: arch_teardown_msi_irq() > expects to find the msi_chip in the chip_data field, which is okay for > all ARM PCI host drivers, but not for other arch MSI chips. > > You fix this by completely removing arch_teardown_msi_irq() at the end > of the series, but this still has the potential to introduce issues for > other arches than ARM within the series. So this patch should include a > change to replace the line > > struct msi_chip *chip = irq_get_chip_data(irq); > > with something that doesn't rely on the msi_chip being in the irq > chip_data field. OK, I will update arch_teardown_msi_irq() in this patch, thanks! Thanks! Yijing. > > Regards, > Lucas >
diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c index f6cb317..d547f7f 100644 --- a/drivers/pci/msi.c +++ b/drivers/pci/msi.c @@ -41,8 +41,6 @@ int __weak arch_setup_msi_irq(struct pci_dev *dev, struct msi_desc *desc) if (err < 0) return err; - irq_set_chip_data(desc->irq, chip); - return 0; }
Currently, pcie-designware, pcie-rcar, pci-tegra drivers use irq chip_data to save the msi_chip pointer. They already call irq_set_chip_data() in their own MSI irq map functions. So irq_set_chip_data() in arch_setup_msi_irq() is useless. Signed-off-by: Yijing Wang <wangyijing@huawei.com> --- drivers/pci/msi.c | 2 -- 1 files changed, 0 insertions(+), 2 deletions(-)