IB/mthca: Add module parameter for number of MTTs per segment

The current MTT allocator uses kmalloc() to allocate a buffer for its
buddy allocator, and thus is limited in the amount of MTT segments
that it can control.  As a result, the size of memory that can be
registered is limited too.  This patch uses a module parameter to
control the number of MTT entries that each segment represents,
allowing more memory to be registered with the same number of
segments.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

5 files changed, 26 insertions, 14 deletions

diff --git a/drivers/infiniband/hw/mthca/mthca_cmd.c b/drivers/infiniband/hw/mthca/mthca_cmd.c
index 6d55f9d..8c2ed99 100644
--- a/drivers/infiniband/hw/mthca/mthca_cmd.c
+++ b/drivers/infiniband/hw/mthca/mthca_cmd.c
@@ -1059,7 +1059,7 @@ int mthca_QUERY_DEV_LIM(struct mthca_dev *dev,
 	MTHCA_GET(field, outbox, QUERY_DEV_LIM_RSVD_MTT_OFFSET);
 	if (mthca_is_memfree(dev))
 		dev_lim->reserved_mtts = ALIGN((1 << (field >> 4)) * sizeof(u64),
-					       MTHCA_MTT_SEG_SIZE) / MTHCA_MTT_SEG_SIZE;
+					       dev->limits.mtt_seg_size) / dev->limits.mtt_seg_size;
 	else
 		dev_lim->reserved_mtts = 1 << (field >> 4);
 	MTHCA_GET(field, outbox, QUERY_DEV_LIM_MAX_MRW_SZ_OFFSET);
diff --git a/drivers/infiniband/hw/mthca/mthca_dev.h b/drivers/infiniband/hw/mthca/mthca_dev.h
index 2525901..9ef611f 100644
--- a/drivers/infiniband/hw/mthca/mthca_dev.h
+++ b/drivers/infiniband/hw/mthca/mthca_dev.h
@@ -159,6 +159,7 @@ struct mthca_limits {
 	int      reserved_eqs;
 	int      num_mpts;
 	int      num_mtt_segs;
+	int	 mtt_seg_size;
 	int      fmr_reserved_mtts;
 	int      reserved_mtts;
 	int      reserved_mrws;
diff --git a/drivers/infiniband/hw/mthca/mthca_main.c b/drivers/infiniband/hw/mthca/mthca_main.c
index 1d83cf7..13da9f1 100644
--- a/drivers/infiniband/hw/mthca/mthca_main.c
+++ b/drivers/infiniband/hw/mthca/mthca_main.c
@@ -125,6 +125,10 @@ module_param_named(fmr_reserved_mtts, hca_profile.fmr_reserved_mtts, int, 0444);
 MODULE_PARM_DESC(fmr_reserved_mtts,
 		 "number of memory translation table segments reserved for FMR");
 
+static int log_mtts_per_seg = ilog2(MTHCA_MTT_SEG_SIZE / 8);
+module_param_named(log_mtts_per_seg, log_mtts_per_seg, int, 0444);
+MODULE_PARM_DESC(log_mtts_per_seg, "Log2 number of MTT entries per segment (1-5)");
+
 static char mthca_version[] __devinitdata =
 	DRV_NAME ": Mellanox InfiniBand HCA driver v"
 	DRV_VERSION " (" DRV_RELDATE ")\n";
@@ -162,6 +166,7 @@ static int mthca_dev_lim(struct mthca_dev *mdev, struct mthca_dev_lim *dev_lim)
 	int err;
 	u8 status;
 
+	mdev->limits.mtt_seg_size = (1 << log_mtts_per_seg) * 8;
 	err = mthca_QUERY_DEV_LIM(mdev, dev_lim, &status);
 	if (err) {
 		mthca_err(mdev, "QUERY_DEV_LIM command failed, aborting.\n");
@@ -460,11 +465,11 @@ static int mthca_init_icm(struct mthca_dev *mdev,
 	}
 
 	/* CPU writes to non-reserved MTTs, while HCA might DMA to reserved mtts */
-	mdev->limits.reserved_mtts = ALIGN(mdev->limits.reserved_mtts * MTHCA_MTT_SEG_SIZE,
-					   dma_get_cache_alignment()) / MTHCA_MTT_SEG_SIZE;
+	mdev->limits.reserved_mtts = ALIGN(mdev->limits.reserved_mtts * mdev->limits.mtt_seg_size,
+					   dma_get_cache_alignment()) / mdev->limits.mtt_seg_size;
 
 	mdev->mr_table.mtt_table = mthca_alloc_icm_table(mdev, init_hca->mtt_base,
-							 MTHCA_MTT_SEG_SIZE,
+							 mdev->limits.mtt_seg_size,
 							 mdev->limits.num_mtt_segs,
 							 mdev->limits.reserved_mtts,
 							 1, 0);
@@ -1315,6 +1320,12 @@ static void __init mthca_validate_profile(void)
 		printk(KERN_WARNING PFX "Corrected fmr_reserved_mtts to %d.\n",
 		       hca_profile.fmr_reserved_mtts);
 	}
+
+	if ((log_mtts_per_seg < 1) || (log_mtts_per_seg > 5)) {
+		printk(KERN_WARNING PFX "bad log_mtts_per_seg (%d). Using default - %d\n",
+		       log_mtts_per_seg, ilog2(MTHCA_MTT_SEG_SIZE / 8));
+		log_mtts_per_seg = ilog2(MTHCA_MTT_SEG_SIZE / 8);
+	}
 }
 
 static int __init mthca_init(void)
diff --git a/drivers/infiniband/hw/mthca/mthca_mr.c b/drivers/infiniband/hw/mthca/mthca_mr.c
index 882e6b7..d606edf 100644
--- a/drivers/infiniband/hw/mthca/mthca_mr.c
+++ b/drivers/infiniband/hw/mthca/mthca_mr.c
@@ -220,7 +220,7 @@ static struct mthca_mtt *__mthca_alloc_mtt(struct mthca_dev *dev, int size,
 
 	mtt->buddy = buddy;
 	mtt->order = 0;
-	for (i = MTHCA_MTT_SEG_SIZE / 8; i < size; i <<= 1)
+	for (i = dev->limits.mtt_seg_size / 8; i < size; i <<= 1)
 		++mtt->order;
 
 	mtt->first_seg = mthca_alloc_mtt_range(dev, mtt->order, buddy);
@@ -267,7 +267,7 @@ static int __mthca_write_mtt(struct mthca_dev *dev, struct mthca_mtt *mtt,
 
 	while (list_len > 0) {
 		mtt_entry[0] = cpu_to_be64(dev->mr_table.mtt_base +
-					   mtt->first_seg * MTHCA_MTT_SEG_SIZE +
+					   mtt->first_seg * dev->limits.mtt_seg_size +
 					   start_index * 8);
 		mtt_entry[1] = 0;
 		for (i = 0; i < list_len && i < MTHCA_MAILBOX_SIZE / 8 - 2; ++i)
@@ -326,7 +326,7 @@ static void mthca_tavor_write_mtt_seg(struct mthca_dev *dev,
 	u64 __iomem *mtts;
 	int i;
 
-	mtts = dev->mr_table.tavor_fmr.mtt_base + mtt->first_seg * MTHCA_MTT_SEG_SIZE +
+	mtts = dev->mr_table.tavor_fmr.mtt_base + mtt->first_seg * dev->limits.mtt_seg_size +
 		start_index * sizeof (u64);
 	for (i = 0; i < list_len; ++i)
 		mthca_write64_raw(cpu_to_be64(buffer_list[i] | MTHCA_MTT_FLAG_PRESENT),
@@ -345,10 +345,10 @@ static void mthca_arbel_write_mtt_seg(struct mthca_dev *dev,
 	/* For Arbel, all MTTs must fit in the same page. */
 	BUG_ON(s / PAGE_SIZE != (s + list_len * sizeof(u64) - 1) / PAGE_SIZE);
 	/* Require full segments */
-	BUG_ON(s % MTHCA_MTT_SEG_SIZE);
+	BUG_ON(s % dev->limits.mtt_seg_size);
 
 	mtts = mthca_table_find(dev->mr_table.mtt_table, mtt->first_seg +
-				s / MTHCA_MTT_SEG_SIZE, &dma_handle);
+				s / dev->limits.mtt_seg_size, &dma_handle);
 
 	BUG_ON(!mtts);
 
@@ -479,7 +479,7 @@ int mthca_mr_alloc(struct mthca_dev *dev, u32 pd, int buffer_size_shift,
 	if (mr->mtt)
 		mpt_entry->mtt_seg =
 			cpu_to_be64(dev->mr_table.mtt_base +
-				    mr->mtt->first_seg * MTHCA_MTT_SEG_SIZE);
+				    mr->mtt->first_seg * dev->limits.mtt_seg_size);
 
 	if (0) {
 		mthca_dbg(dev, "Dumping MPT entry %08x:\n", mr->ibmr.lkey);
@@ -626,7 +626,7 @@ int mthca_fmr_alloc(struct mthca_dev *dev, u32 pd,
 		goto err_out_table;
 	}
 
-	mtt_seg = mr->mtt->first_seg * MTHCA_MTT_SEG_SIZE;
+	mtt_seg = mr->mtt->first_seg * dev->limits.mtt_seg_size;
 
 	if (mthca_is_memfree(dev)) {
 		mr->mem.arbel.mtts = mthca_table_find(dev->mr_table.mtt_table,
@@ -908,7 +908,7 @@ int mthca_init_mr_table(struct mthca_dev *dev)
 			 dev->mr_table.mtt_base);
 
 		dev->mr_table.tavor_fmr.mtt_base =
-			ioremap(addr, mtts * MTHCA_MTT_SEG_SIZE);
+			ioremap(addr, mtts * dev->limits.mtt_seg_size);
 		if (!dev->mr_table.tavor_fmr.mtt_base) {
 			mthca_warn(dev, "MTT ioremap for FMR failed.\n");
 			err = -ENOMEM;
diff --git a/drivers/infiniband/hw/mthca/mthca_profile.c b/drivers/infiniband/hw/mthca/mthca_profile.c
index d168c25..8edb28a 100644
--- a/drivers/infiniband/hw/mthca/mthca_profile.c
+++ b/drivers/infiniband/hw/mthca/mthca_profile.c
@@ -94,7 +94,7 @@ s64 mthca_make_profile(struct mthca_dev *dev,
 	profile[MTHCA_RES_RDB].size  = MTHCA_RDB_ENTRY_SIZE;
 	profile[MTHCA_RES_MCG].size  = MTHCA_MGM_ENTRY_SIZE;
 	profile[MTHCA_RES_MPT].size  = dev_lim->mpt_entry_sz;
-	profile[MTHCA_RES_MTT].size  = MTHCA_MTT_SEG_SIZE;
+	profile[MTHCA_RES_MTT].size  = dev->limits.mtt_seg_size;
 	profile[MTHCA_RES_UAR].size  = dev_lim->uar_scratch_entry_sz;
 	profile[MTHCA_RES_UDAV].size = MTHCA_AV_SIZE;
 	profile[MTHCA_RES_UARC].size = request->uarc_size;
@@ -232,7 +232,7 @@ s64 mthca_make_profile(struct mthca_dev *dev,
 			dev->limits.num_mtt_segs = profile[i].num;
 			dev->mr_table.mtt_base   = profile[i].start;
 			init_hca->mtt_base       = profile[i].start;
-			init_hca->mtt_seg_sz     = ffs(MTHCA_MTT_SEG_SIZE) - 7;
+			init_hca->mtt_seg_sz     = ffs(dev->limits.mtt_seg_size) - 7;
 			break;
 		case MTHCA_RES_UAR:
 			dev->limits.num_uars       = profile[i].num;