diff options
Diffstat (limited to 'drivers/net/ethernet/mellanox/mlxsw/spectrum_buffers.c')
| -rw-r--r-- | drivers/net/ethernet/mellanox/mlxsw/spectrum_buffers.c | 377 |
1 files changed, 348 insertions, 29 deletions
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_buffers.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_buffers.c index 6f84557a5a6f..37ff29a1686e 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_buffers.c +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_buffers.c @@ -121,6 +121,10 @@ struct mlxsw_sp_sb_vals { unsigned int cms_cpu_count; }; +struct mlxsw_sp_sb_ops { + u32 (*int_buf_size_get)(int mtu, u32 speed); +}; + u32 mlxsw_sp_cells_bytes(const struct mlxsw_sp *mlxsw_sp, u32 cells) { return mlxsw_sp->sb->cell_size * cells; @@ -131,9 +135,14 @@ u32 mlxsw_sp_bytes_cells(const struct mlxsw_sp *mlxsw_sp, u32 bytes) return DIV_ROUND_UP(bytes, mlxsw_sp->sb->cell_size); } -u32 mlxsw_sp_sb_max_headroom_cells(const struct mlxsw_sp *mlxsw_sp) +static u32 mlxsw_sp_port_headroom_8x_adjust(const struct mlxsw_sp_port *mlxsw_sp_port, + u32 size_cells) { - return mlxsw_sp->sb->max_headroom_cells; + /* Ports with eight lanes use two headroom buffers between which the + * configured headroom size is split. Therefore, multiply the calculated + * headroom size by two. + */ + return mlxsw_sp_port->mapping.width == 8 ? 2 * size_cells : size_cells; } static struct mlxsw_sp_sb_pr *mlxsw_sp_sb_pr_get(struct mlxsw_sp *mlxsw_sp, @@ -291,55 +300,308 @@ static int mlxsw_sp_sb_pm_occ_query(struct mlxsw_sp *mlxsw_sp, u8 local_port, (unsigned long) pm); } -/* 1/4 of a headroom necessary for 100Gbps port and 100m cable. */ -#define MLXSW_SP_PB_HEADROOM 25632 +void mlxsw_sp_hdroom_prios_reset_buf_idx(struct mlxsw_sp_hdroom *hdroom) +{ + int prio; + + for (prio = 0; prio < IEEE_8021QAZ_MAX_TCS; prio++) { + switch (hdroom->mode) { + case MLXSW_SP_HDROOM_MODE_DCB: + hdroom->prios.prio[prio].buf_idx = hdroom->prios.prio[prio].ets_buf_idx; + break; + case MLXSW_SP_HDROOM_MODE_TC: + hdroom->prios.prio[prio].buf_idx = hdroom->prios.prio[prio].set_buf_idx; + break; + } + } +} + +void mlxsw_sp_hdroom_bufs_reset_lossiness(struct mlxsw_sp_hdroom *hdroom) +{ + int prio; + int i; + + for (i = 0; i < DCBX_MAX_BUFFERS; i++) + hdroom->bufs.buf[i].lossy = true; + + for (prio = 0; prio < IEEE_8021Q_MAX_PRIORITIES; prio++) { + if (!hdroom->prios.prio[prio].lossy) + hdroom->bufs.buf[hdroom->prios.prio[prio].buf_idx].lossy = false; + } +} + +static u16 mlxsw_sp_hdroom_buf_threshold_get(const struct mlxsw_sp *mlxsw_sp, int mtu) +{ + return 2 * mlxsw_sp_bytes_cells(mlxsw_sp, mtu); +} + +static void mlxsw_sp_hdroom_buf_pack(char *pbmc_pl, int index, u16 size, u16 thres, bool lossy) +{ + if (lossy) + mlxsw_reg_pbmc_lossy_buffer_pack(pbmc_pl, index, size); + else + mlxsw_reg_pbmc_lossless_buffer_pack(pbmc_pl, index, size, + thres); +} + +static u16 mlxsw_sp_hdroom_buf_delay_get(const struct mlxsw_sp *mlxsw_sp, + const struct mlxsw_sp_hdroom *hdroom) +{ + u16 delay_cells; + + delay_cells = mlxsw_sp_bytes_cells(mlxsw_sp, hdroom->delay_bytes); + + /* In the worst case scenario the delay will be made up of packets that + * are all of size CELL_SIZE + 1, which means each packet will require + * almost twice its true size when buffered in the switch. We therefore + * multiply this value by the "cell factor", which is close to 2. + * + * Another MTU is added in case the transmitting host already started + * transmitting a maximum length frame when the PFC packet was received. + */ + return 2 * delay_cells + mlxsw_sp_bytes_cells(mlxsw_sp, hdroom->mtu); +} + +static u32 mlxsw_sp_hdroom_int_buf_size_get(struct mlxsw_sp *mlxsw_sp, int mtu, u32 speed) +{ + u32 buffsize = mlxsw_sp->sb_ops->int_buf_size_get(speed, mtu); + + return mlxsw_sp_bytes_cells(mlxsw_sp, buffsize) + 1; +} + +static bool mlxsw_sp_hdroom_buf_is_used(const struct mlxsw_sp_hdroom *hdroom, int buf) +{ + int prio; + + for (prio = 0; prio < IEEE_8021QAZ_MAX_TCS; prio++) { + if (hdroom->prios.prio[prio].buf_idx == buf) + return true; + } + return false; +} + +void mlxsw_sp_hdroom_bufs_reset_sizes(struct mlxsw_sp_port *mlxsw_sp_port, + struct mlxsw_sp_hdroom *hdroom) +{ + struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp; + u16 reserve_cells; + int i; + + /* Internal buffer. */ + reserve_cells = mlxsw_sp_hdroom_int_buf_size_get(mlxsw_sp, mlxsw_sp_port->max_speed, + mlxsw_sp_port->max_mtu); + reserve_cells = mlxsw_sp_port_headroom_8x_adjust(mlxsw_sp_port, reserve_cells); + hdroom->int_buf.reserve_cells = reserve_cells; + + if (hdroom->int_buf.enable) + hdroom->int_buf.size_cells = reserve_cells; + else + hdroom->int_buf.size_cells = 0; + + /* PG buffers. */ + for (i = 0; i < DCBX_MAX_BUFFERS; i++) { + struct mlxsw_sp_hdroom_buf *buf = &hdroom->bufs.buf[i]; + u16 thres_cells; + u16 delay_cells; + + if (!mlxsw_sp_hdroom_buf_is_used(hdroom, i)) { + thres_cells = 0; + delay_cells = 0; + } else if (buf->lossy) { + thres_cells = mlxsw_sp_hdroom_buf_threshold_get(mlxsw_sp, hdroom->mtu); + delay_cells = 0; + } else { + thres_cells = mlxsw_sp_hdroom_buf_threshold_get(mlxsw_sp, hdroom->mtu); + delay_cells = mlxsw_sp_hdroom_buf_delay_get(mlxsw_sp, hdroom); + } + + thres_cells = mlxsw_sp_port_headroom_8x_adjust(mlxsw_sp_port, thres_cells); + delay_cells = mlxsw_sp_port_headroom_8x_adjust(mlxsw_sp_port, delay_cells); + + buf->thres_cells = thres_cells; + if (hdroom->mode == MLXSW_SP_HDROOM_MODE_DCB) { + buf->size_cells = thres_cells + delay_cells; + } else { + /* Do not allow going below the minimum size, even if + * the user requested it. + */ + buf->size_cells = max(buf->set_size_cells, buf->thres_cells); + } + } +} + #define MLXSW_SP_PB_UNUSED 8 -static int mlxsw_sp_port_pb_init(struct mlxsw_sp_port *mlxsw_sp_port) +static int mlxsw_sp_hdroom_configure_buffers(struct mlxsw_sp_port *mlxsw_sp_port, + const struct mlxsw_sp_hdroom *hdroom, bool force) { - const u32 pbs[] = { - [0] = MLXSW_SP_PB_HEADROOM * mlxsw_sp_port->mapping.width, - [9] = MLXSW_PORT_MAX_MTU, - }; struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp; char pbmc_pl[MLXSW_REG_PBMC_LEN]; + bool dirty; + int err; int i; - mlxsw_reg_pbmc_pack(pbmc_pl, mlxsw_sp_port->local_port, - 0xffff, 0xffff / 2); - for (i = 0; i < ARRAY_SIZE(pbs); i++) { - u16 size = mlxsw_sp_bytes_cells(mlxsw_sp, pbs[i]); + dirty = memcmp(&mlxsw_sp_port->hdroom->bufs, &hdroom->bufs, sizeof(hdroom->bufs)); + if (!dirty && !force) + return 0; + + mlxsw_reg_pbmc_pack(pbmc_pl, mlxsw_sp_port->local_port, 0xffff, 0xffff / 2); + for (i = 0; i < MLXSW_SP_PB_COUNT; i++) { + const struct mlxsw_sp_hdroom_buf *buf = &hdroom->bufs.buf[i]; if (i == MLXSW_SP_PB_UNUSED) continue; - size = mlxsw_sp_port_headroom_8x_adjust(mlxsw_sp_port, size); - mlxsw_reg_pbmc_lossy_buffer_pack(pbmc_pl, i, size); + + mlxsw_sp_hdroom_buf_pack(pbmc_pl, i, buf->size_cells, buf->thres_cells, buf->lossy); } - mlxsw_reg_pbmc_lossy_buffer_pack(pbmc_pl, - MLXSW_REG_PBMC_PORT_SHARED_BUF_IDX, 0); - return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(pbmc), pbmc_pl); + + mlxsw_reg_pbmc_lossy_buffer_pack(pbmc_pl, MLXSW_REG_PBMC_PORT_SHARED_BUF_IDX, 0); + err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(pbmc), pbmc_pl); + if (err) + return err; + + mlxsw_sp_port->hdroom->bufs = hdroom->bufs; + return 0; } -static int mlxsw_sp_port_pb_prio_init(struct mlxsw_sp_port *mlxsw_sp_port) +static int mlxsw_sp_hdroom_configure_priomap(struct mlxsw_sp_port *mlxsw_sp_port, + const struct mlxsw_sp_hdroom *hdroom, bool force) { char pptb_pl[MLXSW_REG_PPTB_LEN]; - int i; + bool dirty; + int prio; + int err; + + dirty = memcmp(&mlxsw_sp_port->hdroom->prios, &hdroom->prios, sizeof(hdroom->prios)); + if (!dirty && !force) + return 0; mlxsw_reg_pptb_pack(pptb_pl, mlxsw_sp_port->local_port); - for (i = 0; i < IEEE_8021QAZ_MAX_TCS; i++) - mlxsw_reg_pptb_prio_to_buff_pack(pptb_pl, i, 0); - return mlxsw_reg_write(mlxsw_sp_port->mlxsw_sp->core, MLXSW_REG(pptb), - pptb_pl); + for (prio = 0; prio < IEEE_8021QAZ_MAX_TCS; prio++) + mlxsw_reg_pptb_prio_to_buff_pack(pptb_pl, prio, hdroom->prios.prio[prio].buf_idx); + + err = mlxsw_reg_write(mlxsw_sp_port->mlxsw_sp->core, MLXSW_REG(pptb), pptb_pl); + if (err) + return err; + + mlxsw_sp_port->hdroom->prios = hdroom->prios; + return 0; } -static int mlxsw_sp_port_headroom_init(struct mlxsw_sp_port *mlxsw_sp_port) +static int mlxsw_sp_hdroom_configure_int_buf(struct mlxsw_sp_port *mlxsw_sp_port, + const struct mlxsw_sp_hdroom *hdroom, bool force) +{ + char sbib_pl[MLXSW_REG_SBIB_LEN]; + bool dirty; + int err; + + dirty = memcmp(&mlxsw_sp_port->hdroom->int_buf, &hdroom->int_buf, sizeof(hdroom->int_buf)); + if (!dirty && !force) + return 0; + + mlxsw_reg_sbib_pack(sbib_pl, mlxsw_sp_port->local_port, hdroom->int_buf.size_cells); + err = mlxsw_reg_write(mlxsw_sp_port->mlxsw_sp->core, MLXSW_REG(sbib), sbib_pl); + if (err) + return err; + + mlxsw_sp_port->hdroom->int_buf = hdroom->int_buf; + return 0; +} + +static bool mlxsw_sp_hdroom_bufs_fit(struct mlxsw_sp *mlxsw_sp, + const struct mlxsw_sp_hdroom *hdroom) { + u32 taken_headroom_cells = 0; + int i; + + for (i = 0; i < MLXSW_SP_PB_COUNT; i++) + taken_headroom_cells += hdroom->bufs.buf[i].size_cells; + + taken_headroom_cells += hdroom->int_buf.reserve_cells; + return taken_headroom_cells <= mlxsw_sp->sb->max_headroom_cells; +} + +static int __mlxsw_sp_hdroom_configure(struct mlxsw_sp_port *mlxsw_sp_port, + const struct mlxsw_sp_hdroom *hdroom, bool force) +{ + struct mlxsw_sp_hdroom orig_hdroom; + struct mlxsw_sp_hdroom tmp_hdroom; int err; + int i; + + /* Port buffers need to be configured in three steps. First, all buffers + * with non-zero size are configured. Then, prio-to-buffer map is + * updated, allowing traffic to flow to the now non-zero buffers. + * Finally, zero-sized buffers are configured, because now no traffic + * should be directed to them anymore. This way, in a non-congested + * system, no packet drops are introduced by the reconfiguration. + */ - err = mlxsw_sp_port_pb_init(mlxsw_sp_port); + orig_hdroom = *mlxsw_sp_port->hdroom; + tmp_hdroom = orig_hdroom; + for (i = 0; i < MLXSW_SP_PB_COUNT; i++) { + if (hdroom->bufs.buf[i].size_cells) + tmp_hdroom.bufs.buf[i] = hdroom->bufs.buf[i]; + } + + if (!mlxsw_sp_hdroom_bufs_fit(mlxsw_sp_port->mlxsw_sp, &tmp_hdroom) || + !mlxsw_sp_hdroom_bufs_fit(mlxsw_sp_port->mlxsw_sp, hdroom)) + return -ENOBUFS; + + err = mlxsw_sp_hdroom_configure_buffers(mlxsw_sp_port, &tmp_hdroom, force); if (err) return err; - return mlxsw_sp_port_pb_prio_init(mlxsw_sp_port); + + err = mlxsw_sp_hdroom_configure_priomap(mlxsw_sp_port, hdroom, force); + if (err) + goto err_configure_priomap; + + err = mlxsw_sp_hdroom_configure_buffers(mlxsw_sp_port, hdroom, false); + if (err) + goto err_configure_buffers; + + err = mlxsw_sp_hdroom_configure_int_buf(mlxsw_sp_port, hdroom, false); + if (err) + goto err_configure_int_buf; + + *mlxsw_sp_port->hdroom = *hdroom; + return 0; + +err_configure_int_buf: + mlxsw_sp_hdroom_configure_buffers(mlxsw_sp_port, &tmp_hdroom, false); +err_configure_buffers: + mlxsw_sp_hdroom_configure_priomap(mlxsw_sp_port, &tmp_hdroom, false); +err_configure_priomap: + mlxsw_sp_hdroom_configure_buffers(mlxsw_sp_port, &orig_hdroom, false); + return err; +} + +int mlxsw_sp_hdroom_configure(struct mlxsw_sp_port *mlxsw_sp_port, + const struct mlxsw_sp_hdroom *hdroom) +{ + return __mlxsw_sp_hdroom_configure(mlxsw_sp_port, hdroom, false); +} + +static int mlxsw_sp_port_headroom_init(struct mlxsw_sp_port *mlxsw_sp_port) +{ + struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp; + struct mlxsw_sp_hdroom hdroom = {}; + u32 size9; + int prio; + + hdroom.mtu = mlxsw_sp_port->dev->mtu; + hdroom.mode = MLXSW_SP_HDROOM_MODE_DCB; + for (prio = 0; prio < IEEE_8021QAZ_MAX_TCS; prio++) + hdroom.prios.prio[prio].lossy = true; + + mlxsw_sp_hdroom_bufs_reset_lossiness(&hdroom); + mlxsw_sp_hdroom_bufs_reset_sizes(mlxsw_sp_port, &hdroom); + + /* Buffer 9 is used for control traffic. */ + size9 = mlxsw_sp_port_headroom_8x_adjust(mlxsw_sp_port, mlxsw_sp_port->max_mtu); + hdroom.bufs.buf[9].size_cells = mlxsw_sp_bytes_cells(mlxsw_sp, size9); + + return __mlxsw_sp_hdroom_configure(mlxsw_sp_port, &hdroom, true); } static int mlxsw_sp_sb_port_init(struct mlxsw_sp *mlxsw_sp, @@ -916,6 +1178,46 @@ const struct mlxsw_sp_sb_vals mlxsw_sp2_sb_vals = { .cms_cpu_count = ARRAY_SIZE(mlxsw_sp_cpu_port_sb_cms), }; +static u32 mlxsw_sp1_pb_int_buf_size_get(int mtu, u32 speed) +{ + return mtu * 5 / 2; +} + +static u32 __mlxsw_sp_pb_int_buf_size_get(int mtu, u32 speed, u32 buffer_factor) +{ + return 3 * mtu + buffer_factor * speed / 1000; +} + +#define MLXSW_SP2_SPAN_EG_MIRROR_BUFFER_FACTOR 38 + +static u32 mlxsw_sp2_pb_int_buf_size_get(int mtu, u32 speed) +{ + int factor = MLXSW_SP2_SPAN_EG_MIRROR_BUFFER_FACTOR; + + return __mlxsw_sp_pb_int_buf_size_get(mtu, speed, factor); +} + +#define MLXSW_SP3_SPAN_EG_MIRROR_BUFFER_FACTOR 50 + +static u32 mlxsw_sp3_pb_int_buf_size_get(int mtu, u32 speed) +{ + int factor = MLXSW_SP3_SPAN_EG_MIRROR_BUFFER_FACTOR; + + return __mlxsw_sp_pb_int_buf_size_get(mtu, speed, factor); +} + +const struct mlxsw_sp_sb_ops mlxsw_sp1_sb_ops = { + .int_buf_size_get = mlxsw_sp1_pb_int_buf_size_get, +}; + +const struct mlxsw_sp_sb_ops mlxsw_sp2_sb_ops = { + .int_buf_size_get = mlxsw_sp2_pb_int_buf_size_get, +}; + +const struct mlxsw_sp_sb_ops mlxsw_sp3_sb_ops = { + .int_buf_size_get = mlxsw_sp3_pb_int_buf_size_get, +}; + int mlxsw_sp_buffers_init(struct mlxsw_sp *mlxsw_sp) { u32 max_headroom_size; @@ -995,17 +1297,34 @@ int mlxsw_sp_port_buffers_init(struct mlxsw_sp_port *mlxsw_sp_port) { int err; + mlxsw_sp_port->hdroom = kzalloc(sizeof(*mlxsw_sp_port->hdroom), GFP_KERNEL); + if (!mlxsw_sp_port->hdroom) + return -ENOMEM; + mlxsw_sp_port->hdroom->mtu = mlxsw_sp_port->dev->mtu; + err = mlxsw_sp_port_headroom_init(mlxsw_sp_port); if (err) - return err; + goto err_headroom_init; err = mlxsw_sp_port_sb_cms_init(mlxsw_sp_port); if (err) - return err; + goto err_port_sb_cms_init; err = mlxsw_sp_port_sb_pms_init(mlxsw_sp_port); + if (err) + goto err_port_sb_pms_init; + return 0; +err_port_sb_pms_init: +err_port_sb_cms_init: +err_headroom_init: + kfree(mlxsw_sp_port->hdroom); return err; } +void mlxsw_sp_port_buffers_fini(struct mlxsw_sp_port *mlxsw_sp_port) +{ + kfree(mlxsw_sp_port->hdroom); +} + int mlxsw_sp_sb_pool_get(struct mlxsw_core *mlxsw_core, unsigned int sb_index, u16 pool_index, struct devlink_sb_pool_info *pool_info) |