[PATCH net 0/2] mlx4_core fixes for 5.0-rc

[PATCH net 0/2] mlx4_core fixes for 5.0-rc

[Tariq Toukan]

Hi Dave,

This patchset includes two fixes for the mlx4_core driver.

First patch by Aya fixes inaccurate parsing of some FW fields, mistakenly
including additional (mostly reserved) bits.

Second patch by Jack fixes a wrong (yet harmless) error handling of
calls to copy_to_user() during the CQs init stage.


Series generated against net commit:
49a57857aeea Linux 5.0-rc3

Thanks,
Tariq.


Aya Levin (1):
  net/mlx4_core: Add masking for a few queries on HCA caps

Jack Morgenstein (1):
  net/mlx4_core: Fix error handling when initializing CQ bufs in the
    driver

 drivers/net/ethernet/mellanox/mlx4/cq.c |  6 ++-
 drivers/net/ethernet/mellanox/mlx4/fw.c | 75 ++++++++++++++++++++-------------
 2 files changed, 50 insertions(+), 31 deletions(-)

-- 
1.8.3.1

[PATCH net 1/2] net/mlx4_core: Add masking for a few queries on HCA caps

[Tariq Toukan]

From: Aya Levin <ayal@mellanox.com>

Driver reads the query HCA capabilities without the corresponding masks.
Without the correct masks, the base addresses of the queues are
unaligned.  In addition some reserved bits were wrongly read.  Using the
correct masks, ensures alignment of the base addresses and allows future
firmware versions safe use of the reserved bits.

Fixes: ab9c17a009ee ("mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet")
Fixes: 0ff1fb654bec ("{NET, IB}/mlx4: Add device managed flow steering firmware API")
Signed-off-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx4/fw.c | 75 ++++++++++++++++++++-------------
 1 file changed, 46 insertions(+), 29 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/fw.c b/drivers/net/ethernet/mellanox/mlx4/fw.c
index 7df728f1e5b5..6e501af0e532 100644
--- a/drivers/net/ethernet/mellanox/mlx4/fw.c
+++ b/drivers/net/ethernet/mellanox/mlx4/fw.c
@@ -2067,9 +2067,11 @@ int mlx4_QUERY_HCA(struct mlx4_dev *dev,
 {
 	struct mlx4_cmd_mailbox *mailbox;
 	__be32 *outbox;
+	u64 qword_field;
 	u32 dword_field;
-	int err;
+	u16 word_field;
 	u8 byte_field;
+	int err;
 	static const u8 a0_dmfs_query_hw_steering[] =  {
 		[0] = MLX4_STEERING_DMFS_A0_DEFAULT,
 		[1] = MLX4_STEERING_DMFS_A0_DYNAMIC,
@@ -2097,19 +2099,32 @@ int mlx4_QUERY_HCA(struct mlx4_dev *dev,
 
 	/* QPC/EEC/CQC/EQC/RDMARC attributes */
 
-	MLX4_GET(param->qpc_base,      outbox, INIT_HCA_QPC_BASE_OFFSET);
-	MLX4_GET(param->log_num_qps,   outbox, INIT_HCA_LOG_QP_OFFSET);
-	MLX4_GET(param->srqc_base,     outbox, INIT_HCA_SRQC_BASE_OFFSET);
-	MLX4_GET(param->log_num_srqs,  outbox, INIT_HCA_LOG_SRQ_OFFSET);
-	MLX4_GET(param->cqc_base,      outbox, INIT_HCA_CQC_BASE_OFFSET);
-	MLX4_GET(param->log_num_cqs,   outbox, INIT_HCA_LOG_CQ_OFFSET);
-	MLX4_GET(param->altc_base,     outbox, INIT_HCA_ALTC_BASE_OFFSET);
-	MLX4_GET(param->auxc_base,     outbox, INIT_HCA_AUXC_BASE_OFFSET);
-	MLX4_GET(param->eqc_base,      outbox, INIT_HCA_EQC_BASE_OFFSET);
-	MLX4_GET(param->log_num_eqs,   outbox, INIT_HCA_LOG_EQ_OFFSET);
-	MLX4_GET(param->num_sys_eqs,   outbox, INIT_HCA_NUM_SYS_EQS_OFFSET);
-	MLX4_GET(param->rdmarc_base,   outbox, INIT_HCA_RDMARC_BASE_OFFSET);
-	MLX4_GET(param->log_rd_per_qp, outbox, INIT_HCA_LOG_RD_OFFSET);
+	MLX4_GET(qword_field, outbox, INIT_HCA_QPC_BASE_OFFSET);
+	param->qpc_base = qword_field & ~((u64)0x1f);
+	MLX4_GET(byte_field, outbox, INIT_HCA_LOG_QP_OFFSET);
+	param->log_num_qps = byte_field & 0x1f;
+	MLX4_GET(qword_field, outbox, INIT_HCA_SRQC_BASE_OFFSET);
+	param->srqc_base = qword_field & ~((u64)0x1f);
+	MLX4_GET(byte_field, outbox, INIT_HCA_LOG_SRQ_OFFSET);
+	param->log_num_srqs = byte_field & 0x1f;
+	MLX4_GET(qword_field, outbox, INIT_HCA_CQC_BASE_OFFSET);
+	param->cqc_base = qword_field & ~((u64)0x1f);
+	MLX4_GET(byte_field, outbox, INIT_HCA_LOG_CQ_OFFSET);
+	param->log_num_cqs = byte_field & 0x1f;
+	MLX4_GET(qword_field, outbox, INIT_HCA_ALTC_BASE_OFFSET);
+	param->altc_base = qword_field;
+	MLX4_GET(qword_field, outbox, INIT_HCA_AUXC_BASE_OFFSET);
+	param->auxc_base = qword_field;
+	MLX4_GET(qword_field, outbox, INIT_HCA_EQC_BASE_OFFSET);
+	param->eqc_base = qword_field & ~((u64)0x1f);
+	MLX4_GET(byte_field, outbox, INIT_HCA_LOG_EQ_OFFSET);
+	param->log_num_eqs = byte_field & 0x1f;
+	MLX4_GET(word_field, outbox, INIT_HCA_NUM_SYS_EQS_OFFSET);
+	param->num_sys_eqs = word_field & 0xfff;
+	MLX4_GET(qword_field, outbox, INIT_HCA_RDMARC_BASE_OFFSET);
+	param->rdmarc_base = qword_field & ~((u64)0x1f);
+	MLX4_GET(byte_field, outbox, INIT_HCA_LOG_RD_OFFSET);
+	param->log_rd_per_qp = byte_field & 0x7;
 
 	MLX4_GET(dword_field, outbox, INIT_HCA_FLAGS_OFFSET);
 	if (dword_field & (1 << INIT_HCA_DEVICE_MANAGED_FLOW_STEERING_EN)) {
@@ -2128,22 +2143,21 @@ int mlx4_QUERY_HCA(struct mlx4_dev *dev,
 	/* steering attributes */
 	if (param->steering_mode == MLX4_STEERING_MODE_DEVICE_MANAGED) {
 		MLX4_GET(param->mc_base, outbox, INIT_HCA_FS_BASE_OFFSET);
-		MLX4_GET(param->log_mc_entry_sz, outbox,
-			 INIT_HCA_FS_LOG_ENTRY_SZ_OFFSET);
-		MLX4_GET(param->log_mc_table_sz, outbox,
-			 INIT_HCA_FS_LOG_TABLE_SZ_OFFSET);
-		MLX4_GET(byte_field, outbox,
-			 INIT_HCA_FS_A0_OFFSET);
+		MLX4_GET(byte_field, outbox, INIT_HCA_FS_LOG_ENTRY_SZ_OFFSET);
+		param->log_mc_entry_sz = byte_field & 0x1f;
+		MLX4_GET(byte_field, outbox, INIT_HCA_FS_LOG_TABLE_SZ_OFFSET);
+		param->log_mc_table_sz = byte_field & 0x1f;
+		MLX4_GET(byte_field, outbox, INIT_HCA_FS_A0_OFFSET);
 		param->dmfs_high_steer_mode =
 			a0_dmfs_query_hw_steering[(byte_field >> 6) & 3];
 	} else {
 		MLX4_GET(param->mc_base, outbox, INIT_HCA_MC_BASE_OFFSET);
-		MLX4_GET(param->log_mc_entry_sz, outbox,
-			 INIT_HCA_LOG_MC_ENTRY_SZ_OFFSET);
-		MLX4_GET(param->log_mc_hash_sz,  outbox,
-			 INIT_HCA_LOG_MC_HASH_SZ_OFFSET);
-		MLX4_GET(param->log_mc_table_sz, outbox,
-			 INIT_HCA_LOG_MC_TABLE_SZ_OFFSET);
+		MLX4_GET(byte_field, outbox, INIT_HCA_LOG_MC_ENTRY_SZ_OFFSET);
+		param->log_mc_entry_sz = byte_field & 0x1f;
+		MLX4_GET(byte_field,  outbox, INIT_HCA_LOG_MC_HASH_SZ_OFFSET);
+		param->log_mc_hash_sz = byte_field & 0x1f;
+		MLX4_GET(byte_field, outbox, INIT_HCA_LOG_MC_TABLE_SZ_OFFSET);
+		param->log_mc_table_sz = byte_field & 0x1f;
 	}
 
 	/* CX3 is capable of extending CQEs/EQEs from 32 to 64 bytes */
@@ -2167,15 +2181,18 @@ int mlx4_QUERY_HCA(struct mlx4_dev *dev,
 	/* TPT attributes */
 
 	MLX4_GET(param->dmpt_base,  outbox, INIT_HCA_DMPT_BASE_OFFSET);
-	MLX4_GET(param->mw_enabled, outbox, INIT_HCA_TPT_MW_OFFSET);
-	MLX4_GET(param->log_mpt_sz, outbox, INIT_HCA_LOG_MPT_SZ_OFFSET);
+	MLX4_GET(byte_field, outbox, INIT_HCA_TPT_MW_OFFSET);
+	param->mw_enabled = byte_field >> 7;
+	MLX4_GET(byte_field, outbox, INIT_HCA_LOG_MPT_SZ_OFFSET);
+	param->log_mpt_sz = byte_field & 0x3f;
 	MLX4_GET(param->mtt_base,   outbox, INIT_HCA_MTT_BASE_OFFSET);
 	MLX4_GET(param->cmpt_base,  outbox, INIT_HCA_CMPT_BASE_OFFSET);
 
 	/* UAR attributes */
 
 	MLX4_GET(param->uar_page_sz, outbox, INIT_HCA_UAR_PAGE_SZ_OFFSET);
-	MLX4_GET(param->log_uar_sz, outbox, INIT_HCA_LOG_UAR_SZ_OFFSET);
+	MLX4_GET(byte_field, outbox, INIT_HCA_LOG_UAR_SZ_OFFSET);
+	param->log_uar_sz = byte_field & 0xf;
 
 	/* phv_check enable */
 	MLX4_GET(byte_field, outbox, INIT_HCA_CACHELINE_SZ_OFFSET);
-- 
1.8.3.1

[PATCH net 2/2] net/mlx4_core: Fix error handling when initializing CQ bufs in the driver

[Tariq Toukan]

From: Jack Morgenstein <jackm@dev.mellanox.co.il>

Procedure mlx4_init_user_cqes() handles returns by copy_to_user
incorrectly. copy_to_user() returns the number of bytes not copied.
Thus, a non-zero return should be treated as a -EFAULT error
(as is done elsewhere in the kernel). However, mlx4_init_user_cqes()
error handling simply returns the number of bytes not copied
(instead of -EFAULT).

Note, though, that this is a harmless bug: procedure mlx4_alloc_cq()
(which is the only caller of mlx4_init_user_cqes()) treats any
non-zero return as an error, but that returned error value is processed
internally, and not passed further up the call stack.

In addition, fixes the following sparse warning:
warning: incorrect type in argument 1 (different address spaces)
   expected void [noderef] <asn:1>*to
   got void *buf

Fixes: e45678973dcb ("{net, IB}/mlx4: Initialize CQ buffers in the driver when possible")
Reported by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx4/cq.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/cq.c b/drivers/net/ethernet/mellanox/mlx4/cq.c
index db909b6069b5..65f8a4b6ed0c 100644
--- a/drivers/net/ethernet/mellanox/mlx4/cq.c
+++ b/drivers/net/ethernet/mellanox/mlx4/cq.c
@@ -306,14 +306,16 @@ static int mlx4_init_user_cqes(void *buf, int entries, int cqe_size)
 
 	if (entries_per_copy < entries) {
 		for (i = 0; i < entries / entries_per_copy; i++) {
-			err = copy_to_user(buf, init_ents, PAGE_SIZE);
+			err = copy_to_user((void __user *)buf, init_ents, PAGE_SIZE) ?
+				-EFAULT : 0;
 			if (err)
 				goto out;
 
 			buf += PAGE_SIZE;
 		}
 	} else {
-		err = copy_to_user(buf, init_ents, entries * cqe_size);
+		err = copy_to_user((void __user *)buf, init_ents, entries * cqe_size) ?
+			-EFAULT : 0;
 	}
 
 out:
-- 
1.8.3.1

Re: [PATCH net 0/2] mlx4_core fixes for 5.0-rc

[David Miller]

From: Tariq Toukan <tariqt@mellanox.com>
Date: Tue, 22 Jan 2019 15:19:43 +0200

> This patchset includes two fixes for the mlx4_core driver.
> 
> First patch by Aya fixes inaccurate parsing of some FW fields, mistakenly
> including additional (mostly reserved) bits.
> 
> Second patch by Jack fixes a wrong (yet harmless) error handling of
> calls to copy_to_user() during the CQs init stage.
> 
> 
> Series generated against net commit:
> 49a57857aeea Linux 5.0-rc3

Series applied, and patch #1 queued up for -stable.