| Message ID | 1470661541-26270-2-git-send-email-tn@semihalf.com (mailing list archive) |
|---|---|
| State | New, archived |
| Delegated to: | Bjorn Helgaas |
| Headers | show |
On Mon, Aug 08, 2016 at 03:05:37PM +0200, Tomasz Nowicki wrote: > pci_config_window keeps pointer to pci_ecam_ops and every time > we want to deallocate pci_config_window (pci_ecam_free()) we need to make > sure to free pci_ecam_ops in case it was dynamically allocated prior to > pci_ecam_create() call. I think this is a theoretical problem, right? All the current callers pass a pointer to a static pci_ecam_ops struct that does not need to be deallocated. I see that the next patch uses a pci_ecam_ops struct on the stack, which would definitely cause issues without this patch, so I agree we need something like this; I'm just trying to make the changelog make more sense. > To avoid that extra effort, embed pci_ecam_ops in pci_config_window, > instead of just a pointer. > > Signed-off-by: Tomasz Nowicki <tn@semihalf.com> > To: Jayachandran C <jchandra@broadcom.com> This looks like an editing mistake? > --- > arch/arm64/kernel/pci.c | 2 +- > drivers/pci/ecam.c | 6 +++--- > include/linux/pci-ecam.h | 2 +- > 3 files changed, 5 insertions(+), 5 deletions(-) > > diff --git a/arch/arm64/kernel/pci.c b/arch/arm64/kernel/pci.c > index acf3872..981e828 100644 > --- a/arch/arm64/kernel/pci.c > +++ b/arch/arm64/kernel/pci.c > @@ -183,7 +183,7 @@ struct pci_bus *pci_acpi_scan_root(struct acpi_pci_root *root) > return NULL; > } > > - acpi_pci_root_ops.pci_ops = &ri->cfg->ops->pci_ops; > + acpi_pci_root_ops.pci_ops = &ri->cfg->ops.pci_ops; > bus = acpi_pci_root_create(root, &acpi_pci_root_ops, &ri->common, > ri->cfg); > if (!bus) > diff --git a/drivers/pci/ecam.c b/drivers/pci/ecam.c > index 43ed08d..24242f0 100644 > --- a/drivers/pci/ecam.c > +++ b/drivers/pci/ecam.c > @@ -52,7 +52,7 @@ struct pci_config_window *pci_ecam_create(struct device *dev, > return ERR_PTR(-ENOMEM); > > cfg->parent = dev; > - cfg->ops = ops; > + cfg->ops = *ops; > cfg->busr.start = busr->start; > cfg->busr.end = busr->end; > cfg->busr.flags = IORESOURCE_BUS; > @@ -138,7 +138,7 @@ void __iomem *pci_ecam_map_bus(struct pci_bus *bus, unsigned int devfn, > int where) > { > struct pci_config_window *cfg = bus->sysdata; > - unsigned int devfn_shift = cfg->ops->bus_shift - 8; > + unsigned int devfn_shift = cfg->ops.bus_shift - 8; > unsigned int busn = bus->number; > void __iomem *base; > > @@ -149,7 +149,7 @@ void __iomem *pci_ecam_map_bus(struct pci_bus *bus, unsigned int devfn, > if (per_bus_mapping) > base = cfg->winp[busn]; > else > - base = cfg->win + (busn << cfg->ops->bus_shift); > + base = cfg->win + (busn << cfg->ops.bus_shift); > return base + (devfn << devfn_shift) + where; > } > > diff --git a/include/linux/pci-ecam.h b/include/linux/pci-ecam.h > index 7adad20..0ce2920 100644 > --- a/include/linux/pci-ecam.h > +++ b/include/linux/pci-ecam.h > @@ -39,7 +39,7 @@ struct pci_config_window { > struct resource res; > struct resource busr; > void *priv; > - struct pci_ecam_ops *ops; > + struct pci_ecam_ops ops; > union { > void __iomem *win; /* 64-bit single mapping */ > void __iomem **winp; /* 32-bit per-bus mapping */ > -- > 1.9.1 > -- 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 Thu, Sep 01, 2016 at 01:23:45PM -0500, Bjorn Helgaas wrote: > On Mon, Aug 08, 2016 at 03:05:37PM +0200, Tomasz Nowicki wrote: > > pci_config_window keeps pointer to pci_ecam_ops and every time > > we want to deallocate pci_config_window (pci_ecam_free()) we need to make > > sure to free pci_ecam_ops in case it was dynamically allocated prior to > > pci_ecam_create() call. > > I think this is a theoretical problem, right? All the current callers > pass a pointer to a static pci_ecam_ops struct that does not need to > be deallocated. > > I see that the next patch uses a pci_ecam_ops struct on the stack, I asked Tomasz why we need to have pci_ecam_ops on the stack in the first place since I do not think it is needed, or nicer (actually I think it is not nice at all, what's the problem in making it static ?). Anyway, that's part of this thread but I think Tomasz was looking more for feedback/agreement on the whole quirks mechanism to check if it is fine, more below. > which would definitely cause issues without this patch, so I agree we > need something like this; I'm just trying to make the changelog make > more sense. > > > To avoid that extra effort, embed pci_ecam_ops in pci_config_window, > > instead of just a pointer. > > > > Signed-off-by: Tomasz Nowicki <tn@semihalf.com> > > To: Jayachandran C <jchandra@broadcom.com> > > This looks like an editing mistake? Yes it is. I think the most important bit about this series is to agree on whether it is ok to keep quirks in drivers/pci/host/mcfg-quirks.c or we need to find a different dir for it, and of course to make sure the quirks mechanism is fine with everyone. Thanks ! Lorenzo > > arch/arm64/kernel/pci.c | 2 +- > > drivers/pci/ecam.c | 6 +++--- > > include/linux/pci-ecam.h | 2 +- > > 3 files changed, 5 insertions(+), 5 deletions(-) > > > > diff --git a/arch/arm64/kernel/pci.c b/arch/arm64/kernel/pci.c > > index acf3872..981e828 100644 > > --- a/arch/arm64/kernel/pci.c > > +++ b/arch/arm64/kernel/pci.c > > @@ -183,7 +183,7 @@ struct pci_bus *pci_acpi_scan_root(struct acpi_pci_root *root) > > return NULL; > > } > > > > - acpi_pci_root_ops.pci_ops = &ri->cfg->ops->pci_ops; > > + acpi_pci_root_ops.pci_ops = &ri->cfg->ops.pci_ops; > > bus = acpi_pci_root_create(root, &acpi_pci_root_ops, &ri->common, > > ri->cfg); > > if (!bus) > > diff --git a/drivers/pci/ecam.c b/drivers/pci/ecam.c > > index 43ed08d..24242f0 100644 > > --- a/drivers/pci/ecam.c > > +++ b/drivers/pci/ecam.c > > @@ -52,7 +52,7 @@ struct pci_config_window *pci_ecam_create(struct device *dev, > > return ERR_PTR(-ENOMEM); > > > > cfg->parent = dev; > > - cfg->ops = ops; > > + cfg->ops = *ops; > > cfg->busr.start = busr->start; > > cfg->busr.end = busr->end; > > cfg->busr.flags = IORESOURCE_BUS; > > @@ -138,7 +138,7 @@ void __iomem *pci_ecam_map_bus(struct pci_bus *bus, unsigned int devfn, > > int where) > > { > > struct pci_config_window *cfg = bus->sysdata; > > - unsigned int devfn_shift = cfg->ops->bus_shift - 8; > > + unsigned int devfn_shift = cfg->ops.bus_shift - 8; > > unsigned int busn = bus->number; > > void __iomem *base; > > > > @@ -149,7 +149,7 @@ void __iomem *pci_ecam_map_bus(struct pci_bus *bus, unsigned int devfn, > > if (per_bus_mapping) > > base = cfg->winp[busn]; > > else > > - base = cfg->win + (busn << cfg->ops->bus_shift); > > + base = cfg->win + (busn << cfg->ops.bus_shift); > > return base + (devfn << devfn_shift) + where; > > } > > > > diff --git a/include/linux/pci-ecam.h b/include/linux/pci-ecam.h > > index 7adad20..0ce2920 100644 > > --- a/include/linux/pci-ecam.h > > +++ b/include/linux/pci-ecam.h > > @@ -39,7 +39,7 @@ struct pci_config_window { > > struct resource res; > > struct resource busr; > > void *priv; > > - struct pci_ecam_ops *ops; > > + struct pci_ecam_ops ops; > > union { > > void __iomem *win; /* 64-bit single mapping */ > > void __iomem **winp; /* 32-bit per-bus mapping */ > > -- > > 1.9.1 > > > -- 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 Fri, Sep 02, 2016 at 04:38:45PM +0100, Lorenzo Pieralisi wrote: > On Thu, Sep 01, 2016 at 01:23:45PM -0500, Bjorn Helgaas wrote: > > On Mon, Aug 08, 2016 at 03:05:37PM +0200, Tomasz Nowicki wrote: > > > pci_config_window keeps pointer to pci_ecam_ops and every time > > > we want to deallocate pci_config_window (pci_ecam_free()) we need to make > > > sure to free pci_ecam_ops in case it was dynamically allocated prior to > > > pci_ecam_create() call. > > > > I think this is a theoretical problem, right? All the current callers > > pass a pointer to a static pci_ecam_ops struct that does not need to > > be deallocated. > > > > I see that the next patch uses a pci_ecam_ops struct on the stack, > > I asked Tomasz why we need to have pci_ecam_ops on the stack in the > first place since I do not think it is needed, or nicer (actually I > think it is not nice at all, what's the problem in making it static ?). If it's useful to copy the struct instead of saving the pointer, that's fine. I just want the changelog to match the diff. Bjorn -- 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/arch/arm64/kernel/pci.c b/arch/arm64/kernel/pci.c index acf3872..981e828 100644 --- a/arch/arm64/kernel/pci.c +++ b/arch/arm64/kernel/pci.c @@ -183,7 +183,7 @@ struct pci_bus *pci_acpi_scan_root(struct acpi_pci_root *root) return NULL; } - acpi_pci_root_ops.pci_ops = &ri->cfg->ops->pci_ops; + acpi_pci_root_ops.pci_ops = &ri->cfg->ops.pci_ops; bus = acpi_pci_root_create(root, &acpi_pci_root_ops, &ri->common, ri->cfg); if (!bus) diff --git a/drivers/pci/ecam.c b/drivers/pci/ecam.c index 43ed08d..24242f0 100644 --- a/drivers/pci/ecam.c +++ b/drivers/pci/ecam.c @@ -52,7 +52,7 @@ struct pci_config_window *pci_ecam_create(struct device *dev, return ERR_PTR(-ENOMEM); cfg->parent = dev; - cfg->ops = ops; + cfg->ops = *ops; cfg->busr.start = busr->start; cfg->busr.end = busr->end; cfg->busr.flags = IORESOURCE_BUS; @@ -138,7 +138,7 @@ void __iomem *pci_ecam_map_bus(struct pci_bus *bus, unsigned int devfn, int where) { struct pci_config_window *cfg = bus->sysdata; - unsigned int devfn_shift = cfg->ops->bus_shift - 8; + unsigned int devfn_shift = cfg->ops.bus_shift - 8; unsigned int busn = bus->number; void __iomem *base; @@ -149,7 +149,7 @@ void __iomem *pci_ecam_map_bus(struct pci_bus *bus, unsigned int devfn, if (per_bus_mapping) base = cfg->winp[busn]; else - base = cfg->win + (busn << cfg->ops->bus_shift); + base = cfg->win + (busn << cfg->ops.bus_shift); return base + (devfn << devfn_shift) + where; } diff --git a/include/linux/pci-ecam.h b/include/linux/pci-ecam.h index 7adad20..0ce2920 100644 --- a/include/linux/pci-ecam.h +++ b/include/linux/pci-ecam.h @@ -39,7 +39,7 @@ struct pci_config_window { struct resource res; struct resource busr; void *priv; - struct pci_ecam_ops *ops; + struct pci_ecam_ops ops; union { void __iomem *win; /* 64-bit single mapping */ void __iomem **winp; /* 32-bit per-bus mapping */
pci_config_window keeps pointer to pci_ecam_ops and every time we want to deallocate pci_config_window (pci_ecam_free()) we need to make sure to free pci_ecam_ops in case it was dynamically allocated prior to pci_ecam_create() call. To avoid that extra effort, embed pci_ecam_ops in pci_config_window, instead of just a pointer. Signed-off-by: Tomasz Nowicki <tn@semihalf.com> To: Jayachandran C <jchandra@broadcom.com> --- arch/arm64/kernel/pci.c | 2 +- drivers/pci/ecam.c | 6 +++--- include/linux/pci-ecam.h | 2 +- 3 files changed, 5 insertions(+), 5 deletions(-)