On Mon, Aug 24, 2009 at 03:54:16PM +0000, David Vrabel wrote:
> The WHCI HCD will also support urbs with scatter-gather lists. Add a
> usb_bus field to indicated how many sg list elements are supported by
> the HCD. Use this to decide whether to pass the scatter-list to the HCD
> or not.
>
> Always use scatter-gather urbs where possible, regardless of the
> endpoint type. If the scatter-gather list doesn't have suitably sized
> elements for a particular endpoint type then splitting it up into
> separate urbs won't help.
Does this mean control URBs may have a scatter gather list? The xHCI
driver can't currently handle that.
> Make the usb-storage driver use this new field.
>
> Signed-off-by: David Vrabel <david.vrabel@xxxxxxx>
> Cc: Sarah Sharp <sarah.a.sharp@xxxxxxxxxxxxxxx>
> ---
> Sarah,
>
> Could you check the reasoning in the 2nd paragraph above and confirm
> that the xhci-hcd's sg_tablesize is a suitable value?
>
> drivers/usb/core/message.c | 8 +-------
> drivers/usb/host/xhci-pci.c | 2 ++
> drivers/usb/storage/usb.c | 10 ++++++++++
> include/linux/usb.h | 1 +
> 4 files changed, 14 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/usb/core/message.c b/drivers/usb/core/message.c
> index 9720e69..e17bded 100644
> --- a/drivers/usb/core/message.c
> +++ b/drivers/usb/core/message.c
> @@ -393,13 +393,7 @@ int usb_sg_init(struct usb_sg_request *io, struct usb_device *dev,
> if (io->entries <= 0)
> return io->entries;
>
> - /* If we're running on an xHCI host controller, queue the whole scatter
> - * gather list with one call to urb_enqueue(). This is only for bulk,
> - * as that endpoint type does not care how the data gets broken up
> - * across frames.
> - */
> - if (usb_pipebulk(pipe) &&
> - bus_to_hcd(dev->bus)->driver->flags & HCD_USB3) {
> + if (dev->bus->sg_tablesize > 0) {
> io->urbs = kmalloc(sizeof *io->urbs, mem_flags);
> use_sg = true;
> } else {
> diff --git a/drivers/usb/host/xhci-pci.c b/drivers/usb/host/xhci-pci.c
> index 592fe7e..bbc3934 100644
> --- a/drivers/usb/host/xhci-pci.c
> +++ b/drivers/usb/host/xhci-pci.c
> @@ -50,6 +50,8 @@ static int xhci_pci_setup(struct usb_hcd *hcd)
> struct pci_dev *pdev = to_pci_dev(hcd->self.controller);
> int retval;
>
> + hcd->self.sg_tablesize = 2048;
> +
Please set this to (TRBS_PER_SEGMENT - 1). It's not a very big number
(currently 63). I haven't added dynamic transfer ring resizing yet, so
there's only 63 TRBs to map to sg list entries.
> xhci->cap_regs = hcd->regs;
> xhci->op_regs = hcd->regs +
> HC_LENGTH(xhci_readl(xhci, &xhci->cap_regs->hc_capbase));
> diff --git a/drivers/usb/storage/usb.c b/drivers/usb/storage/usb.c
> index 8060b85..964d8ee 100644
> --- a/drivers/usb/storage/usb.c
> +++ b/drivers/usb/storage/usb.c
> @@ -840,6 +840,15 @@ static int usb_stor_scan_thread(void * __us)
> complete_and_exit(&us->scanning_done, 0);
> }
>
> +static unsigned int usb_stor_sg_tablesize(struct usb_interface *intf)
> +{
> + struct usb_device *usb_dev = interface_to_usbdev(intf);
> +
> + if (usb_dev->bus->sg_tablesize) {
> + return usb_dev->bus->sg_tablesize;
> + }
> + return SG_ALL;
> +}
>
> /* First part of general USB mass-storage probing */
> int usb_stor_probe1(struct us_data **pus,
> @@ -868,6 +877,7 @@ int usb_stor_probe1(struct us_data **pus,
> * Allow 16-byte CDBs and thus > 2TB
> */
> host->max_cmd_len = 16;
> + host->sg_tablesize = usb_stor_sg_tablesize(intf);
> *pus = us = host_to_us(host);
> memset(us, 0, sizeof(struct us_data));
> mutex_init(&(us->dev_mutex));
> diff --git a/include/linux/usb.h b/include/linux/usb.h
> index b1e3c2f..199e467 100644
> --- a/include/linux/usb.h
> +++ b/include/linux/usb.h
> @@ -331,6 +331,7 @@ struct usb_bus {
> u8 otg_port; /* 0, or number of OTG/HNP port */
> unsigned is_b_host:1; /* true during some HNP roleswitches */
> unsigned b_hnp_enable:1; /* OTG: did A-Host enable HNP? */
> + unsigned sg_tablesize; /* 0 or largest number of sg list entries */
>
> int devnum_next; /* Next open device number in
> * round-robin allocation */
> --
> 1.6.3.3
>
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
