* [PATCH] mcb: mcb-pci: Only remap the 1st 0x200 bytes of BAR 0
@ 2014-10-28 16:11 Johannes Thumshirn
2014-10-28 16:22 ` Bjorn Helgaas
2014-12-15 13:23 ` Johannes Thumshirn
0 siblings, 2 replies; 4+ messages in thread
From: Johannes Thumshirn @ 2014-10-28 16:11 UTC (permalink / raw)
To: Greg Kroah-Hartman
Cc: Bjorn Helgaas, Andreas Werner, linux-kernel, Johannes Thumshirn
Currently it is not possible to have a kernel with built-in MCB attached
devices. This results out of the fact that mcb-pci requests PCI BAR 0, then
parses the chameleon table and calls the driver's probe function before
releasing BAR 0 again. When building the kernel with modules this is not a
problem (and therefore it wasn't detected by my tests yet).
A solution is to only remap the 1st 0x200 bytes of a Chameleon PCI device.
0x200 bytes is the maximum size of a Chameleon v2 Table.
Also this patch stops disabling the PCI device on successful registration of MCB
devices.
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@men.de>
Suggested-by: Bjorn Helgaas <bhelgaas@google.com>
---
drivers/mcb/mcb-internal.h | 1 +
drivers/mcb/mcb-pci.c | 27 ++++++++++++++++++---------
2 files changed, 19 insertions(+), 9 deletions(-)
diff --git a/drivers/mcb/mcb-internal.h b/drivers/mcb/mcb-internal.h
index f956ef2..fb7493d 100644
--- a/drivers/mcb/mcb-internal.h
+++ b/drivers/mcb/mcb-internal.h
@@ -7,6 +7,7 @@
#define PCI_DEVICE_ID_MEN_CHAMELEON 0x4d45
#define CHAMELEON_FILENAME_LEN 12
#define CHAMELEONV2_MAGIC 0xabce
+#define CHAM_HEADER_SIZE 0x200
enum chameleon_descriptor_type {
CHAMELEON_DTYPE_GENERAL = 0x0,
diff --git a/drivers/mcb/mcb-pci.c b/drivers/mcb/mcb-pci.c
index b591819..5e1bd5d 100644
--- a/drivers/mcb/mcb-pci.c
+++ b/drivers/mcb/mcb-pci.c
@@ -17,6 +17,7 @@
struct priv {
struct mcb_bus *bus;
+ phys_addr_t mapbase;
void __iomem *base;
};
@@ -31,8 +32,8 @@ static int mcb_pci_get_irq(struct mcb_device *mdev)
static int mcb_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
{
+ struct resource *res;
struct priv *priv;
- phys_addr_t mapbase;
int ret;
int num_cells;
unsigned long flags;
@@ -47,19 +48,21 @@ static int mcb_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
return -ENODEV;
}
- mapbase = pci_resource_start(pdev, 0);
- if (!mapbase) {
+ priv->mapbase = pci_resource_start(pdev, 0);
+ if (!priv->mapbase) {
dev_err(&pdev->dev, "No PCI resource\n");
goto err_start;
}
- ret = pci_request_region(pdev, 0, KBUILD_MODNAME);
- if (ret) {
- dev_err(&pdev->dev, "Failed to request PCI BARs\n");
+ res = request_mem_region(priv->mapbase, CHAM_HEADER_SIZE,
+ KBUILD_MODNAME);
+ if (IS_ERR(res)) {
+ dev_err(&pdev->dev, "Failed to request PCI memory\n");
+ ret = PTR_ERR(res);
goto err_start;
}
- priv->base = pci_iomap(pdev, 0, 0);
+ priv->base = ioremap(priv->mapbase, CHAM_HEADER_SIZE);
if (!priv->base) {
dev_err(&pdev->dev, "Cannot ioremap\n");
ret = -ENOMEM;
@@ -84,7 +87,7 @@ static int mcb_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
priv->bus->get_irq = mcb_pci_get_irq;
- ret = chameleon_parse_cells(priv->bus, mapbase, priv->base);
+ ret = chameleon_parse_cells(priv->bus, priv->mapbase, priv->base);
if (ret < 0)
goto err_drvdata;
num_cells = ret;
@@ -93,8 +96,10 @@ static int mcb_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
mcb_bus_add_devices(priv->bus);
+ return 0;
+
err_drvdata:
- pci_iounmap(pdev, priv->base);
+ iounmap(priv->base);
err_ioremap:
pci_release_region(pdev, 0);
err_start:
@@ -107,6 +112,10 @@ static void mcb_pci_remove(struct pci_dev *pdev)
struct priv *priv = pci_get_drvdata(pdev);
mcb_release_bus(priv->bus);
+
+ iounmap(priv->base);
+ release_region(priv->mapbase, CHAM_HEADER_SIZE);
+ pci_disable_device(pdev);
}
static const struct pci_device_id mcb_pci_tbl[] = {
--
1.9.1
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH] mcb: mcb-pci: Only remap the 1st 0x200 bytes of BAR 0
2014-10-28 16:11 [PATCH] mcb: mcb-pci: Only remap the 1st 0x200 bytes of BAR 0 Johannes Thumshirn
@ 2014-10-28 16:22 ` Bjorn Helgaas
2014-12-15 13:23 ` Johannes Thumshirn
1 sibling, 0 replies; 4+ messages in thread
From: Bjorn Helgaas @ 2014-10-28 16:22 UTC (permalink / raw)
To: Johannes Thumshirn; +Cc: Greg Kroah-Hartman, Andreas Werner, linux-kernel
On Tue, Oct 28, 2014 at 10:11 AM, Johannes Thumshirn
<johannes.thumshirn@men.de> wrote:
> Currently it is not possible to have a kernel with built-in MCB attached
> devices. This results out of the fact that mcb-pci requests PCI BAR 0, then
> parses the chameleon table and calls the driver's probe function before
> releasing BAR 0 again. When building the kernel with modules this is not a
> problem (and therefore it wasn't detected by my tests yet).
>
> A solution is to only remap the 1st 0x200 bytes of a Chameleon PCI device.
> 0x200 bytes is the maximum size of a Chameleon v2 Table.
>
> Also this patch stops disabling the PCI device on successful registration of MCB
> devices.
>
> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@men.de>
> Suggested-by: Bjorn Helgaas <bhelgaas@google.com>
Looks good to me!
Reviewed-by: Bjorn Helgaas <bhelgaas@google.com>
> ---
> drivers/mcb/mcb-internal.h | 1 +
> drivers/mcb/mcb-pci.c | 27 ++++++++++++++++++---------
> 2 files changed, 19 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/mcb/mcb-internal.h b/drivers/mcb/mcb-internal.h
> index f956ef2..fb7493d 100644
> --- a/drivers/mcb/mcb-internal.h
> +++ b/drivers/mcb/mcb-internal.h
> @@ -7,6 +7,7 @@
> #define PCI_DEVICE_ID_MEN_CHAMELEON 0x4d45
> #define CHAMELEON_FILENAME_LEN 12
> #define CHAMELEONV2_MAGIC 0xabce
> +#define CHAM_HEADER_SIZE 0x200
>
> enum chameleon_descriptor_type {
> CHAMELEON_DTYPE_GENERAL = 0x0,
> diff --git a/drivers/mcb/mcb-pci.c b/drivers/mcb/mcb-pci.c
> index b591819..5e1bd5d 100644
> --- a/drivers/mcb/mcb-pci.c
> +++ b/drivers/mcb/mcb-pci.c
> @@ -17,6 +17,7 @@
>
> struct priv {
> struct mcb_bus *bus;
> + phys_addr_t mapbase;
> void __iomem *base;
> };
>
> @@ -31,8 +32,8 @@ static int mcb_pci_get_irq(struct mcb_device *mdev)
>
> static int mcb_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
> {
> + struct resource *res;
> struct priv *priv;
> - phys_addr_t mapbase;
> int ret;
> int num_cells;
> unsigned long flags;
> @@ -47,19 +48,21 @@ static int mcb_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
> return -ENODEV;
> }
>
> - mapbase = pci_resource_start(pdev, 0);
> - if (!mapbase) {
> + priv->mapbase = pci_resource_start(pdev, 0);
> + if (!priv->mapbase) {
> dev_err(&pdev->dev, "No PCI resource\n");
> goto err_start;
> }
>
> - ret = pci_request_region(pdev, 0, KBUILD_MODNAME);
> - if (ret) {
> - dev_err(&pdev->dev, "Failed to request PCI BARs\n");
> + res = request_mem_region(priv->mapbase, CHAM_HEADER_SIZE,
> + KBUILD_MODNAME);
> + if (IS_ERR(res)) {
> + dev_err(&pdev->dev, "Failed to request PCI memory\n");
> + ret = PTR_ERR(res);
> goto err_start;
> }
>
> - priv->base = pci_iomap(pdev, 0, 0);
> + priv->base = ioremap(priv->mapbase, CHAM_HEADER_SIZE);
> if (!priv->base) {
> dev_err(&pdev->dev, "Cannot ioremap\n");
> ret = -ENOMEM;
> @@ -84,7 +87,7 @@ static int mcb_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
>
> priv->bus->get_irq = mcb_pci_get_irq;
>
> - ret = chameleon_parse_cells(priv->bus, mapbase, priv->base);
> + ret = chameleon_parse_cells(priv->bus, priv->mapbase, priv->base);
> if (ret < 0)
> goto err_drvdata;
> num_cells = ret;
> @@ -93,8 +96,10 @@ static int mcb_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
>
> mcb_bus_add_devices(priv->bus);
>
> + return 0;
> +
> err_drvdata:
> - pci_iounmap(pdev, priv->base);
> + iounmap(priv->base);
> err_ioremap:
> pci_release_region(pdev, 0);
> err_start:
> @@ -107,6 +112,10 @@ static void mcb_pci_remove(struct pci_dev *pdev)
> struct priv *priv = pci_get_drvdata(pdev);
>
> mcb_release_bus(priv->bus);
> +
> + iounmap(priv->base);
> + release_region(priv->mapbase, CHAM_HEADER_SIZE);
> + pci_disable_device(pdev);
> }
>
> static const struct pci_device_id mcb_pci_tbl[] = {
> --
> 1.9.1
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] mcb: mcb-pci: Only remap the 1st 0x200 bytes of BAR 0
2014-10-28 16:11 [PATCH] mcb: mcb-pci: Only remap the 1st 0x200 bytes of BAR 0 Johannes Thumshirn
2014-10-28 16:22 ` Bjorn Helgaas
@ 2014-12-15 13:23 ` Johannes Thumshirn
2014-12-15 17:01 ` Greg Kroah-Hartman
1 sibling, 1 reply; 4+ messages in thread
From: Johannes Thumshirn @ 2014-12-15 13:23 UTC (permalink / raw)
To: Greg Kroah-Hartman
Cc: Bjorn Helgaas, Andreas Werner, linux-kernel, Johannes Thumshirn
On Tue, Oct 28, 2014 at 05:11:32PM +0100, Johannes Thumshirn wrote:
> Currently it is not possible to have a kernel with built-in MCB attached
> devices. This results out of the fact that mcb-pci requests PCI BAR 0, then
> parses the chameleon table and calls the driver's probe function before
> releasing BAR 0 again. When building the kernel with modules this is not a
> problem (and therefore it wasn't detected by my tests yet).
>
> A solution is to only remap the 1st 0x200 bytes of a Chameleon PCI device.
> 0x200 bytes is the maximum size of a Chameleon v2 Table.
>
> Also this patch stops disabling the PCI device on successful registration of MCB
> devices.
>
> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@men.de>
> Suggested-by: Bjorn Helgaas <bhelgaas@google.com>
Hi Greg,
Is there any specific reason why you didn't pick this one up for 3.19? Or did it
just got lost in your queue?
Thanks,
Johannes
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] mcb: mcb-pci: Only remap the 1st 0x200 bytes of BAR 0
2014-12-15 13:23 ` Johannes Thumshirn
@ 2014-12-15 17:01 ` Greg Kroah-Hartman
0 siblings, 0 replies; 4+ messages in thread
From: Greg Kroah-Hartman @ 2014-12-15 17:01 UTC (permalink / raw)
To: Johannes Thumshirn; +Cc: Bjorn Helgaas, Andreas Werner, linux-kernel
On Mon, Dec 15, 2014 at 02:23:37PM +0100, Johannes Thumshirn wrote:
> On Tue, Oct 28, 2014 at 05:11:32PM +0100, Johannes Thumshirn wrote:
> > Currently it is not possible to have a kernel with built-in MCB attached
> > devices. This results out of the fact that mcb-pci requests PCI BAR 0, then
> > parses the chameleon table and calls the driver's probe function before
> > releasing BAR 0 again. When building the kernel with modules this is not a
> > problem (and therefore it wasn't detected by my tests yet).
> >
> > A solution is to only remap the 1st 0x200 bytes of a Chameleon PCI device.
> > 0x200 bytes is the maximum size of a Chameleon v2 Table.
> >
> > Also this patch stops disabling the PCI device on successful registration of MCB
> > devices.
> >
> > Signed-off-by: Johannes Thumshirn <johannes.thumshirn@men.de>
> > Suggested-by: Bjorn Helgaas <bhelgaas@google.com>
>
> Hi Greg,
>
> Is there any specific reason why you didn't pick this one up for 3.19? Or did it
> just got lost in your queue?
>
Hm, I didn't realize I was supposed to pick it up, sorry, I'm not listed
as the "mcb" maintainer, you are :)
Please resend and I'll pick it up after 3.19-rc1 is out.
thanks,
greg k-h
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2014-12-15 17:01 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-10-28 16:11 [PATCH] mcb: mcb-pci: Only remap the 1st 0x200 bytes of BAR 0 Johannes Thumshirn
2014-10-28 16:22 ` Bjorn Helgaas
2014-12-15 13:23 ` Johannes Thumshirn
2014-12-15 17:01 ` Greg Kroah-Hartman
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.