Skip to content

Commit

Permalink
bnx2x: fix memory barriers
Browse files Browse the repository at this point in the history 
Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
  • Loading branch information
Vladislav Zolotarov authored and David S. Miller committed Jul 19, 2011
1 parent f5219d8 commit 53e51e2
Show file tree
Hide file tree
Showing 2 changed files with 75 additions and 19 deletions.
8 changes: 6 additions & 2 deletions drivers/net/bnx2x/bnx2x_main.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2984,8 +2984,12 @@ static inline void bnx2x_sp_prod_update(struct bnx2x *bp)
{
int func = BP_FUNC(bp);

/* Make sure that BD data is updated before writing the producer */
wmb();
/*
* Make sure that BD data is updated before writing the producer:
* BD data is written to the memory, the producer is read from the
* memory, thus we need a full memory barrier to ensure the ordering.
*/
mb();

REG_WR16(bp, BAR_XSTRORM_INTMEM + XSTORM_SPQ_PROD_OFFSET(func),
bp->spq_prod_idx);
Expand Down
86 changes: 69 additions & 17 deletions drivers/net/bnx2x/bnx2x_sp.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1563,8 +1563,13 @@ static int bnx2x_execute_vlan_mac(struct bnx2x *bp,
idx++;
}

/* Commit the data writes towards the memory */
mb();
/*
* No need for an explicit memory barrier here as long we would
* need to ensure the ordering of writing to the SPQ element
* and updating of the SPQ producer which involves a memory
* read and we will have to put a full memory barrier there
* (inside bnx2x_sp_post()).
*/

rc = bnx2x_sp_post(bp, o->ramrod_cmd, r->cid,
U64_HI(r->rdata_mapping),
Expand Down Expand Up @@ -2224,8 +2229,13 @@ static int bnx2x_set_rx_mode_e2(struct bnx2x *bp,
data->header.rule_cnt, p->rx_accept_flags,
p->tx_accept_flags);

/* Commit writes towards the memory before sending a ramrod */
mb();
/*
* No need for an explicit memory barrier here as long we would
* need to ensure the ordering of writing to the SPQ element
* and updating of the SPQ producer which involves a memory
* read and we will have to put a full memory barrier there
* (inside bnx2x_sp_post()).
*/

/* Send a ramrod */
rc = bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_FILTER_RULES, p->cid,
Expand Down Expand Up @@ -2918,16 +2928,22 @@ static int bnx2x_mcast_setup_e2(struct bnx2x *bp,
if (!o->total_pending_num)
bnx2x_mcast_refresh_registry_e2(bp, o);

/* Commit writes towards the memory before sending a ramrod */
mb();

/* If CLEAR_ONLY was requested - don't send a ramrod and clear
/*
* If CLEAR_ONLY was requested - don't send a ramrod and clear
* RAMROD_PENDING status immediately.
*/
if (test_bit(RAMROD_DRV_CLR_ONLY, &p->ramrod_flags)) {
raw->clear_pending(raw);
return 0;
} else {
/*
* No need for an explicit memory barrier here as long we would
* need to ensure the ordering of writing to the SPQ element
* and updating of the SPQ producer which involves a memory
* read and we will have to put a full memory barrier there
* (inside bnx2x_sp_post()).
*/

/* Send a ramrod */
rc = bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_MULTICAST_RULES,
raw->cid, U64_HI(raw->rdata_mapping),
Expand Down Expand Up @@ -3404,16 +3420,22 @@ static int bnx2x_mcast_setup_e1(struct bnx2x *bp,
if (rc)
return rc;

/* Commit writes towards the memory before sending a ramrod */
mb();

/* If CLEAR_ONLY was requested - don't send a ramrod and clear
/*
* If CLEAR_ONLY was requested - don't send a ramrod and clear
* RAMROD_PENDING status immediately.
*/
if (test_bit(RAMROD_DRV_CLR_ONLY, &p->ramrod_flags)) {
raw->clear_pending(raw);
return 0;
} else {
/*
* No need for an explicit memory barrier here as long we would
* need to ensure the ordering of writing to the SPQ element
* and updating of the SPQ producer which involves a memory
* read and we will have to put a full memory barrier there
* (inside bnx2x_sp_post()).
*/

/* Send a ramrod */
rc = bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_SET_MAC, raw->cid,
U64_HI(raw->rdata_mapping),
Expand Down Expand Up @@ -4038,8 +4060,13 @@ static int bnx2x_setup_rss(struct bnx2x *bp,
data->capabilities |= ETH_RSS_UPDATE_RAMROD_DATA_UPDATE_RSS_KEY;
}

/* Commit writes towards the memory before sending a ramrod */
mb();
/*
* No need for an explicit memory barrier here as long we would
* need to ensure the ordering of writing to the SPQ element
* and updating of the SPQ producer which involves a memory
* read and we will have to put a full memory barrier there
* (inside bnx2x_sp_post()).
*/

/* Send a ramrod */
rc = bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_RSS_UPDATE, r->cid,
Expand Down Expand Up @@ -4505,7 +4532,13 @@ static inline int bnx2x_q_send_setup_e1x(struct bnx2x *bp,
/* Fill the ramrod data */
bnx2x_q_fill_setup_data_cmn(bp, params, rdata);

mb();
/*
* No need for an explicit memory barrier here as long we would
* need to ensure the ordering of writing to the SPQ element
* and updating of the SPQ producer which involves a memory
* read and we will have to put a full memory barrier there
* (inside bnx2x_sp_post()).
*/

return bnx2x_sp_post(bp, ramrod, o->cids[BNX2X_PRIMARY_CID_INDEX],
U64_HI(data_mapping),
Expand All @@ -4528,6 +4561,13 @@ static inline int bnx2x_q_send_setup_e2(struct bnx2x *bp,
bnx2x_q_fill_setup_data_cmn(bp, params, rdata);
bnx2x_q_fill_setup_data_e2(bp, params, rdata);

/*
* No need for an explicit memory barrier here as long we would
* need to ensure the ordering of writing to the SPQ element
* and updating of the SPQ producer which involves a memory
* read and we will have to put a full memory barrier there
* (inside bnx2x_sp_post()).
*/

return bnx2x_sp_post(bp, ramrod, o->cids[BNX2X_PRIMARY_CID_INDEX],
U64_HI(data_mapping),
Expand Down Expand Up @@ -4665,7 +4705,13 @@ static inline int bnx2x_q_send_update(struct bnx2x *bp,
/* Fill the ramrod data */
bnx2x_q_fill_update_data(bp, o, update_params, rdata);

mb();
/*
* No need for an explicit memory barrier here as long we would
* need to ensure the ordering of writing to the SPQ element
* and updating of the SPQ producer which involves a memory
* read and we will have to put a full memory barrier there
* (inside bnx2x_sp_post()).
*/

return bnx2x_sp_post(bp, RAMROD_CMD_ID_ETH_CLIENT_UPDATE,
o->cids[cid_index], U64_HI(data_mapping),
Expand Down Expand Up @@ -5484,7 +5530,13 @@ static inline int bnx2x_func_send_start(struct bnx2x *bp,
rdata->path_id = BP_PATH(bp);
rdata->network_cos_mode = start_params->network_cos_mode;

mb();
/*
* No need for an explicit memory barrier here as long we would
* need to ensure the ordering of writing to the SPQ element
* and updating of the SPQ producer which involves a memory
* read and we will have to put a full memory barrier there
* (inside bnx2x_sp_post()).
*/

return bnx2x_sp_post(bp, RAMROD_CMD_ID_COMMON_FUNCTION_START, 0,
U64_HI(data_mapping),
Expand Down

0 comments on commit 53e51e2

Please sign in to comment.