From: Alan Stern stern@rowland.harvard.edu
commit 1186f86a71130a7635a20843e355bb880c7349b2 upstream.
Commit 3ae62a42090f ("UAS: fix alignment of scatter/gather segments"), copying a similar commit for usb-storage, attempted to solve a problem involving scatter-gather I/O and USB/IP by setting the virt_boundary_mask for mass-storage devices.
However, it now turns out that the analogous change in usb-storage interacted badly with commit 09324d32d2a0 ("block: force an unlimited segment size on queues with a virt boundary"), which was added later. A typical error message is:
ehci-pci 0000:00:13.2: swiotlb buffer is full (sz: 327680 bytes), total 32768 (slots), used 97 (slots)
There is no longer any reason to keep the virt_boundary_mask setting in the uas driver. It was needed in the first place only for handling devices with a block size smaller than the maxpacket size and where the host controller was not capable of fully general scatter-gather operation (that is, able to merge two SG segments into a single USB packet). But:
High-speed or slower connections never use a bulk maxpacket value larger than 512;
The SCSI layer does not handle block devices with a block size smaller than 512 bytes;
All the host controllers capable of SuperSpeed operation can handle fully general SG;
Since commit ea44d190764b ("usbip: Implement SG support to vhci-hcd and stub driver") was merged, the USB/IP driver can also handle SG.
Therefore all supported device/controller combinations should be okay with no need for any special virt_boundary_mask. So in order to head off potential problems similar to those affecting usb-storage, this patch reverts commit 3ae62a42090f.
Signed-off-by: Alan Stern stern@rowland.harvard.edu CC: Oliver Neukum oneukum@suse.com CC: stable@vger.kernel.org Acked-by: Christoph Hellwig hch@lst.de Fixes: 3ae62a42090f ("UAS: fix alignment of scatter/gather segments") Link: https://lore.kernel.org/r/Pine.LNX.4.44L0.1910231132470.1878-100000@iolanthe... Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org
--- drivers/usb/storage/uas.c | 20 -------------------- 1 file changed, 20 deletions(-)
--- a/drivers/usb/storage/uas.c +++ b/drivers/usb/storage/uas.c @@ -789,30 +789,10 @@ static int uas_slave_alloc(struct scsi_d { struct uas_dev_info *devinfo = (struct uas_dev_info *)sdev->host->hostdata; - int maxp;
sdev->hostdata = devinfo;
/* - * We have two requirements here. We must satisfy the requirements - * of the physical HC and the demands of the protocol, as we - * definitely want no additional memory allocation in this path - * ruling out using bounce buffers. - * - * For a transmission on USB to continue we must never send - * a package that is smaller than maxpacket. Hence the length of each - * scatterlist element except the last must be divisible by the - * Bulk maxpacket value. - * If the HC does not ensure that through SG, - * the upper layer must do that. We must assume nothing - * about the capabilities off the HC, so we use the most - * pessimistic requirement. - */ - - maxp = usb_maxpacket(devinfo->udev, devinfo->data_in_pipe, 0); - blk_queue_virt_boundary(sdev->request_queue, maxp - 1); - - /* * The protocol has no requirements on alignment in the strict sense. * Controllers may or may not have alignment restrictions. * As this is not exported, we use an extremely conservative guess.