From 4c4c0a89abd5c08e91df9bcce4ebcb3433bbb9bf Mon Sep 17 00:00:00 2001 From: Parav Pandit Date: Fri, 19 Feb 2021 08:18:12 +0200 Subject: net/mlx5: Pack mlx5_rl_entry structure mlx5_rl_entry structure is not properly packed as shown below. Due to this an array of size 9144 bytes allocated which is aligned to 16Kbytes. Hence, pack the structure and avoid the wastage. This offers 8Kbytes of saving per mlx5_core_dev struct. pahole -C mlx5_rl_entry drivers/net/ethernet/mellanox/mlx5/core/en_main.o Existing layout: struct mlx5_rl_entry { u8 rl_raw[48]; /* 0 48 */ u16 index; /* 48 2 */ /* XXX 6 bytes hole, try to pack */ u64 refcount; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ u16 uid; /* 64 2 */ u8 dedicated:1; /* 66: 0 1 */ /* size: 72, cachelines: 2, members: 5 */ /* sum members: 60, holes: 1, sum holes: 6 */ /* sum bitfield members: 1 bits (0 bytes) */ /* padding: 5 */ /* bit_padding: 7 bits */ /* last cacheline: 8 bytes */ }; After alignment: struct mlx5_rl_entry { u8 rl_raw[48]; /* 0 48 */ u64 refcount; /* 48 8 */ u16 index; /* 56 2 */ u16 uid; /* 58 2 */ u8 dedicated:1; /* 60: 0 1 */ /* size: 64, cachelines: 1, members: 5 */ /* padding: 3 */ /* bit_padding: 7 bits */ }; Signed-off-by: Parav Pandit Signed-off-by: Saeed Mahameed --- include/linux/mlx5/driver.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include/linux') diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h index 23bb01d7c9b9..a9bd7e3bd554 100644 --- a/include/linux/mlx5/driver.h +++ b/include/linux/mlx5/driver.h @@ -517,8 +517,8 @@ struct mlx5_rate_limit { struct mlx5_rl_entry { u8 rl_raw[MLX5_ST_SZ_BYTES(set_pp_rate_limit_context)]; - u16 index; u64 refcount; + u16 index; u16 uid; u8 dedicated : 1; }; -- cgit v1.2.3 From 6b30b6d4d36c978e0ab0f22e85bf3c646732e98b Mon Sep 17 00:00:00 2001 From: Parav Pandit Date: Fri, 19 Feb 2021 12:06:54 +0200 Subject: net/mlx5: Allocate rate limit table when rate is configured A device supports 128 rate limiters. A static table allocation consumes 8KB of memory even when rate is not configured. Instead, allocate the table when at least one rate is configured. Signed-off-by: Parav Pandit Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/rl.c | 46 ++++++++++++++++++++++------ include/linux/mlx5/driver.h | 1 + 2 files changed, 38 insertions(+), 9 deletions(-) (limited to 'include/linux') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/rl.c b/drivers/net/ethernet/mellanox/mlx5/core/rl.c index 08792fe701e3..0526e3798c09 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/rl.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/rl.c @@ -117,6 +117,9 @@ static struct mlx5_rl_entry *find_rl_entry(struct mlx5_rl_table *table, bool empty_found = false; int i; + lockdep_assert_held(&table->rl_lock); + WARN_ON(!table->rl_entry); + for (i = 0; i < table->max_size; i++) { if (dedicated) { if (!table->rl_entry[i].refcount) @@ -172,10 +175,17 @@ bool mlx5_rl_are_equal(struct mlx5_rate_limit *rl_0, } EXPORT_SYMBOL(mlx5_rl_are_equal); -static int mlx5_rl_table_alloc(struct mlx5_rl_table *table) +static int mlx5_rl_table_get(struct mlx5_rl_table *table) { int i; + lockdep_assert_held(&table->rl_lock); + + if (table->rl_entry) { + table->refcount++; + return 0; + } + table->rl_entry = kcalloc(table->max_size, sizeof(struct mlx5_rl_entry), GFP_KERNEL); if (!table->rl_entry) @@ -187,13 +197,27 @@ static int mlx5_rl_table_alloc(struct mlx5_rl_table *table) for (i = 0; i < table->max_size; i++) table->rl_entry[i].index = i + 1; + table->refcount++; return 0; } +static void mlx5_rl_table_put(struct mlx5_rl_table *table) +{ + lockdep_assert_held(&table->rl_lock); + if (--table->refcount) + return; + + kfree(table->rl_entry); + table->rl_entry = NULL; +} + static void mlx5_rl_table_free(struct mlx5_core_dev *dev, struct mlx5_rl_table *table) { int i; + if (!table->rl_entry) + return; + /* Clear all configured rates */ for (i = 0; i < table->max_size; i++) if (table->rl_entry[i].refcount) @@ -219,8 +243,8 @@ int mlx5_rl_add_rate_raw(struct mlx5_core_dev *dev, void *rl_in, u16 uid, { struct mlx5_rl_table *table = &dev->priv.rl_table; struct mlx5_rl_entry *entry; - int err = 0; u32 rate; + int err; if (!table->max_size) return -EOPNOTSUPP; @@ -233,13 +257,16 @@ int mlx5_rl_add_rate_raw(struct mlx5_core_dev *dev, void *rl_in, u16 uid, } mutex_lock(&table->rl_lock); + err = mlx5_rl_table_get(table); + if (err) + goto out; entry = find_rl_entry(table, rl_in, uid, dedicated_entry); if (!entry) { mlx5_core_err(dev, "Max number of %u rates reached\n", table->max_size); err = -ENOSPC; - goto out; + goto rl_err; } if (!entry->refcount) { /* new rate limit */ @@ -255,14 +282,18 @@ int mlx5_rl_add_rate_raw(struct mlx5_core_dev *dev, void *rl_in, u16 uid, burst_upper_bound), MLX5_GET(set_pp_rate_limit_context, rl_in, typical_packet_size)); - goto out; + goto rl_err; } entry->dedicated = dedicated_entry; } mlx5_rl_entry_get(entry); *index = entry->index; + mutex_unlock(&table->rl_lock); + return 0; +rl_err: + mlx5_rl_table_put(table); out: mutex_unlock(&table->rl_lock); return err; @@ -277,6 +308,7 @@ void mlx5_rl_remove_rate_raw(struct mlx5_core_dev *dev, u16 index) mutex_lock(&table->rl_lock); entry = &table->rl_entry[index - 1]; mlx5_rl_entry_put(dev, entry); + mlx5_rl_table_put(table); mutex_unlock(&table->rl_lock); } EXPORT_SYMBOL(mlx5_rl_remove_rate_raw); @@ -325,6 +357,7 @@ void mlx5_rl_remove_rate(struct mlx5_core_dev *dev, struct mlx5_rate_limit *rl) goto out; } mlx5_rl_entry_put(dev, entry); + mlx5_rl_table_put(table); out: mutex_unlock(&table->rl_lock); } @@ -333,7 +366,6 @@ EXPORT_SYMBOL(mlx5_rl_remove_rate); int mlx5_init_rl_table(struct mlx5_core_dev *dev) { struct mlx5_rl_table *table = &dev->priv.rl_table; - int err; mutex_init(&table->rl_lock); if (!MLX5_CAP_GEN(dev, qos) || !MLX5_CAP_QOS(dev, packet_pacing)) { @@ -346,10 +378,6 @@ int mlx5_init_rl_table(struct mlx5_core_dev *dev) table->max_rate = MLX5_CAP_QOS(dev, packet_pacing_max_rate); table->min_rate = MLX5_CAP_QOS(dev, packet_pacing_min_rate); - err = mlx5_rl_table_alloc(table); - if (err) - return err; - mlx5_core_info(dev, "Rate limit: %u rates are supported, range: %uMbps to %uMbps\n", table->max_size, table->min_rate >> 10, diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h index a9bd7e3bd554..baf38b5a2a8c 100644 --- a/include/linux/mlx5/driver.h +++ b/include/linux/mlx5/driver.h @@ -530,6 +530,7 @@ struct mlx5_rl_table { u32 max_rate; u32 min_rate; struct mlx5_rl_entry *rl_entry; + u64 refcount; }; struct mlx5_core_roce { -- cgit v1.2.3