scsi: core: Add a dma_alignment field to the host and host template

Get drivers out of the business of having to call the block layer DMA
alignment limits helpers themselves.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240409143748.980206-8-hch@lst.deReviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: John Garry <john.g.garry@oracle.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

drivers/firewire/sbp2.c

@@ -1500,12 +1500,6 @@ static int sbp2_scsi_slave_alloc(struct scsi_device *sdev)
 
 	sdev->allow_restart = 1;
 
-	/*
-	 * SBP-2 does not require any alignment, but we set it anyway
-	 * for compatibility with earlier versions of this driver.
-	 */
-	blk_queue_update_dma_alignment(sdev->request_queue, 4 - 1);
-
 	if (lu->tgt->workarounds & SBP2_WORKAROUND_INQUIRY_36)
 		sdev->inquiry_len = 36;
 

drivers/message/fusion/mptfc.c

@@ -129,6 +129,7 @@ static const struct scsi_host_template mptfc_driver_template = {
 	.sg_tablesize			= MPT_SCSI_SG_DEPTH,
 	.max_sectors			= 8192,
 	.cmd_per_lun			= 7,
+	.dma_alignment			= 511,
 	.shost_groups			= mptscsih_host_attr_groups,
 };
 

drivers/message/fusion/mptsas.c

@@ -2020,6 +2020,7 @@ static const struct scsi_host_template mptsas_driver_template = {
 	.sg_tablesize			= MPT_SCSI_SG_DEPTH,
 	.max_sectors			= 8192,
 	.cmd_per_lun			= 7,
+	.dma_alignment			= 511,
 	.shost_groups			= mptscsih_host_attr_groups,
 	.no_write_same			= 1,
 };

drivers/message/fusion/mptscsih.c

@@ -2438,8 +2438,6 @@ mptscsih_slave_configure(struct scsi_device *sdev)
 		"tagged %d, simple %d\n",
 		ioc->name,sdev->tagged_supported, sdev->simple_tags));
 
-	blk_queue_dma_alignment (sdev->request_queue, 512 - 1);
-
 	return 0;
 }
 

drivers/message/fusion/mptspi.c

@@ -843,6 +843,7 @@ static const struct scsi_host_template mptspi_driver_template = {
 	.sg_tablesize			= MPT_SCSI_SG_DEPTH,
 	.max_sectors			= 8192,
 	.cmd_per_lun			= 7,
+	.dma_alignment			= 511,
 	.shost_groups			= mptscsih_host_attr_groups,
 };
 

drivers/scsi/hosts.c

@@ -478,6 +478,12 @@ struct Scsi_Host *scsi_host_alloc(const struct scsi_host_template *sht, int priv
 	else
 		shost->max_segment_size = BLK_MAX_SEGMENT_SIZE;
 
+	/* 32-byte (dword) is a common minimum for HBAs. */
+	if (sht->dma_alignment)
+		shost->dma_alignment = sht->dma_alignment;
+	else
+		shost->dma_alignment = 3;
+
 	/*
 	 * assume a 4GB boundary, if not set
 	 */

drivers/scsi/iscsi_tcp.c

@@ -943,6 +943,7 @@ iscsi_sw_tcp_session_create(struct iscsi_endpoint *ep, uint16_t cmds_max,
 	shost->max_id = 0;
 	shost->max_channel = 0;
 	shost->max_cmd_len = SCSI_MAX_VARLEN_CDB_SIZE;
+	shost->dma_alignment = 0;
 
 	rc = iscsi_host_get_max_scsi_cmds(shost, cmds_max);
 	if (rc < 0)
@@ -1065,7 +1066,6 @@ static int iscsi_sw_tcp_slave_configure(struct scsi_device *sdev)
 	if (conn->datadgst_en)
 		blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES,
 				   sdev->request_queue);
-	blk_queue_dma_alignment(sdev->request_queue, 0);
 	return 0;
 }
 

drivers/scsi/qla2xxx/qla_os.c

@@ -1957,9 +1957,6 @@ qla2xxx_slave_configure(struct scsi_device *sdev)
 	scsi_qla_host_t *vha = shost_priv(sdev->host);
 	struct req_que *req = vha->req;
 
-	if (IS_T10_PI_CAPABLE(vha->hw))
-		blk_queue_update_dma_alignment(sdev->request_queue, 0x7);
-
 	scsi_change_queue_depth(sdev, req->max_q_depth);
 	return 0;
 }
@@ -3575,6 +3572,9 @@ qla2x00_probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
 		    QLA_SG_ALL : 128;
 	}
 
+	if (IS_T10_PI_CAPABLE(base_vha->hw))
+		host->dma_alignment = 0x7;
+
 	ret = scsi_add_host(host, &pdev->dev);
 	if (ret)
 		goto probe_failed;

drivers/scsi/scsi_lib.c

@@ -1985,15 +1985,8 @@ void scsi_init_limits(struct Scsi_Host *shost, struct queue_limits *lim)
 	lim->seg_boundary_mask = shost->dma_boundary;
 	lim->max_segment_size = shost->max_segment_size;
 	lim->virt_boundary_mask = shost->virt_boundary_mask;
-
-	/*
-	 * Set a reasonable default alignment:  The larger of 32-byte (dword),
-	 * which is a common minimum for HBAs, and the minimum DMA alignment,
-	 * which is set by the platform.
-	 *
-	 * Devices that require a bigger alignment can increase it later.
-	 */
-	lim->dma_alignment = max(4, dma_get_cache_alignment()) - 1;
+	lim->dma_alignment = max_t(unsigned int,
+		shost->dma_alignment, dma_get_cache_alignment() - 1);
 
 	if (shost->no_highmem)
 		lim->bounce = BLK_BOUNCE_HIGH;

drivers/staging/rts5208/rtsx.c

@@ -70,18 +70,6 @@ static int slave_alloc(struct scsi_device *sdev)
 
 static int slave_configure(struct scsi_device *sdev)
 {
-	/*
-	 * Scatter-gather buffers (all but the last) must have a length
-	 * divisible by the bulk maxpacket size.  Otherwise a data packet
-	 * would end up being short, causing a premature end to the data
-	 * transfer.  Since high-speed bulk pipes have a maxpacket size
-	 * of 512, we'll use that as the scsi device queue's DMA alignment
-	 * mask.  Guaranteeing proper alignment of the first buffer will
-	 * have the desired effect because, except at the beginning and
-	 * the end, scatter-gather buffers follow page boundaries.
-	 */
-	blk_queue_dma_alignment(sdev->request_queue, (512 - 1));
-
 	/* Set the SCSI level to at least 2.  We'll leave it at 3 if that's
 	 * what is originally reported.  We need this to avoid confusing
 	 * the SCSI layer with devices that report 0 or 1, but need 10-byte
@@ -219,6 +207,18 @@ static const struct scsi_host_template rtsx_host_template = {
 	/* limit the total size of a transfer to 120 KB */
 	.max_sectors =                  240,
 
+	/*
+	 * Scatter-gather buffers (all but the last) must have a length
+	 * divisible by the bulk maxpacket size.  Otherwise a data packet
+	 * would end up being short, causing a premature end to the data
+	 * transfer.  Since high-speed bulk pipes have a maxpacket size
+	 * of 512, we'll use that as the scsi device queue's DMA alignment
+	 * mask.  Guaranteeing proper alignment of the first buffer will
+	 * have the desired effect because, except at the beginning and
+	 * the end, scatter-gather buffers follow page boundaries.
+	 */
+	.dma_alignment =		511,
+
 	/* emulated HBA */
 	.emulated =			1,
 

drivers/usb/image/microtek.c

@@ -328,12 +328,6 @@ static int mts_slave_alloc (struct scsi_device *s)
 	return 0;
 }
 
-static int mts_slave_configure (struct scsi_device *s)
-{
-	blk_queue_dma_alignment(s->request_queue, (512 - 1));
-	return 0;
-}
-
 static int mts_scsi_abort(struct scsi_cmnd *srb)
 {
 	struct mts_desc* desc = (struct mts_desc*)(srb->device->host->hostdata[0]);
@@ -631,8 +625,8 @@ static const struct scsi_host_template mts_scsi_host_template = {
 	.can_queue =		1,
 	.this_id =		-1,
 	.emulated =		1,
+	.dma_alignment =	511,
 	.slave_alloc =		mts_slave_alloc,
-	.slave_configure =	mts_slave_configure,
 	.max_sectors=		256, /* 128 K */
 };
 

drivers/usb/storage/scsiglue.c

@@ -75,12 +75,6 @@ static int slave_alloc (struct scsi_device *sdev)
 	 */
 	sdev->inquiry_len = 36;
 
-	/*
-	 * Some host controllers may have alignment requirements.
-	 * We'll play it safe by requiring 512-byte alignment always.
-	 */
-	blk_queue_update_dma_alignment(sdev->request_queue, (512 - 1));
-
 	/* Tell the SCSI layer if we know there is more than one LUN */
 	if (us->protocol == USB_PR_BULK && us->max_lun > 0)
 		sdev->sdev_bflags |= BLIST_FORCELUN;
@@ -638,6 +632,11 @@ static const struct scsi_host_template usb_stor_host_template = {
 	/* lots of sg segments can be handled */
 	.sg_tablesize =			SG_MAX_SEGMENTS,
 
+	/*
+	 * Some host controllers may have alignment requirements.
+	 * We'll play it safe by requiring 512-byte alignment always.
+	 */
+	.dma_alignment =		511,
 
 	/*
 	 * Limit the total size of a transfer to 120 KB.

drivers/usb/storage/uas.c

@@ -824,13 +824,6 @@ static int uas_slave_alloc(struct scsi_device *sdev)
 
 	sdev->hostdata = devinfo;
 
-	/*
-	 * 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.
-	 */
-	blk_queue_update_dma_alignment(sdev->request_queue, (512 - 1));
-
 	if (devinfo->flags & US_FL_MAX_SECTORS_64)
 		blk_queue_max_hw_sectors(sdev->request_queue, 64);
 	else if (devinfo->flags & US_FL_MAX_SECTORS_240)
@@ -912,6 +905,12 @@ static const struct scsi_host_template uas_host_template = {
 	.eh_device_reset_handler = uas_eh_device_reset_handler,
 	.this_id = -1,
 	.skip_settle_delay = 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.
+	 */
+	.dma_alignment = 511,
 	.dma_boundary = PAGE_SIZE - 1,
 	.cmd_size = sizeof(struct uas_cmd_info),
 };

include/scsi/scsi_host.h

@@ -405,6 +405,8 @@ struct scsi_host_template {
 	 */
 	unsigned int max_segment_size;
 
+	unsigned int dma_alignment;
+
 	/*
 	 * DMA scatter gather segment boundary limit. A segment crossing this
 	 * boundary will be split in two.
@@ -614,6 +616,7 @@ struct Scsi_Host {
 	unsigned int max_sectors;
 	unsigned int opt_sectors;
 	unsigned int max_segment_size;
+	unsigned int dma_alignment;
 	unsigned long dma_boundary;
 	unsigned long virt_boundary_mask;
 	/*