| Message ID | 1724341989-27612-1-git-send-email-shradhagupta@linux.microsoft.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [net-next,v4] net: mana: Implement get_ringparam/set_ringparam for mana | expand |
On Thu, Aug 22, 2024 at 08:53:09AM -0700, Shradha Gupta wrote: > Currently the values of WQs for RX and TX queues for MANA devices > are hardcoded to default sizes. > Allow configuring these values for MANA devices as ringparam > configuration(get/set) through ethtool_ops. > Pre-allocate buffers at the beginning of this operation, to > prevent complete network loss in low-memory conditions. > > Signed-off-by: Shradha Gupta <shradhagupta@linux.microsoft.com> > Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com> > --- > Changes in v4: > * Roundup the ring parameter value to a power of 2 > * Skip the max value check for parameters > * Use extack to log errors > --- > Changes in v3: > * pre-allocate buffers before changing the queue sizes > * rebased to latest net-next > --- > Changes in v2: > * Removed unnecessary validations in mana_set_ringparam() > * Fixed codespell error > * Improved error message to indicate issue with the parameter > --- > drivers/net/ethernet/microsoft/mana/mana_en.c | 24 +++--- > .../ethernet/microsoft/mana/mana_ethtool.c | 74 +++++++++++++++++++ > include/net/mana/mana.h | 23 +++++- > 3 files changed, 108 insertions(+), 13 deletions(-) > > diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c b/drivers/net/ethernet/microsoft/mana/mana_en.c > index d2f07e179e86..4e3ade5926bc 100644 > --- a/drivers/net/ethernet/microsoft/mana/mana_en.c > +++ b/drivers/net/ethernet/microsoft/mana/mana_en.c > @@ -511,7 +511,7 @@ static u16 mana_select_queue(struct net_device *ndev, struct sk_buff *skb, > } > > /* Release pre-allocated RX buffers */ > -static void mana_pre_dealloc_rxbufs(struct mana_port_context *mpc) > +void mana_pre_dealloc_rxbufs(struct mana_port_context *mpc) > { > struct device *dev; > int i; > @@ -604,7 +604,7 @@ static void mana_get_rxbuf_cfg(int mtu, u32 *datasize, u32 *alloc_size, > *datasize = mtu + ETH_HLEN; > } > > -static int mana_pre_alloc_rxbufs(struct mana_port_context *mpc, int new_mtu) > +int mana_pre_alloc_rxbufs(struct mana_port_context *mpc, int new_mtu) > { > struct device *dev; > struct page *page; > @@ -618,7 +618,7 @@ static int mana_pre_alloc_rxbufs(struct mana_port_context *mpc, int new_mtu) > > dev = mpc->ac->gdma_dev->gdma_context->dev; > > - num_rxb = mpc->num_queues * RX_BUFFERS_PER_QUEUE; > + num_rxb = mpc->num_queues * mpc->rx_queue_size; > > WARN(mpc->rxbufs_pre, "mana rxbufs_pre exists\n"); > mpc->rxbufs_pre = kmalloc_array(num_rxb, sizeof(void *), GFP_KERNEL); > @@ -1899,14 +1899,15 @@ static int mana_create_txq(struct mana_port_context *apc, > return -ENOMEM; > > /* The minimum size of the WQE is 32 bytes, hence > - * MAX_SEND_BUFFERS_PER_QUEUE represents the maximum number of WQEs > + * apc->tx_queue_size represents the maximum number of WQEs > * the SQ can store. This value is then used to size other queues > * to prevent overflow. > + * Also note that the txq_size is always going to be MANA_PAGE_ALIGNED, > + * as tx_queue_size is always a power of 2. > */ MANA_PAGE_ALIGNED aligned means aligned by 0x1000. tx_queue_size being 'power of 2' * 32 is not a sufficient condition for it to be aligned to 0x1000. We possibly can explain more. > - txq_size = MAX_SEND_BUFFERS_PER_QUEUE * 32; > - BUILD_BUG_ON(!MANA_PAGE_ALIGNED(txq_size)); > + txq_size = apc->tx_queue_size * 32; > > - cq_size = MAX_SEND_BUFFERS_PER_QUEUE * COMP_ENTRY_SIZE; > + cq_size = apc->tx_queue_size * COMP_ENTRY_SIZE; > cq_size = MANA_PAGE_ALIGN(cq_size); COMP_ENTRY_SIZE is 64, that means cq_size is double of txq_size. If we are certain that txq_size is always aligned to MANA_PAGE, that means cq_size is already aligned to MANA_PAGE as well. - Saurabh
On Fri, Aug 23, 2024 at 04:34:54AM -0700, Saurabh Singh Sengar wrote: > On Thu, Aug 22, 2024 at 08:53:09AM -0700, Shradha Gupta wrote: > > Currently the values of WQs for RX and TX queues for MANA devices > > are hardcoded to default sizes. > > Allow configuring these values for MANA devices as ringparam > > configuration(get/set) through ethtool_ops. > > Pre-allocate buffers at the beginning of this operation, to > > prevent complete network loss in low-memory conditions. > > > > Signed-off-by: Shradha Gupta <shradhagupta@linux.microsoft.com> > > Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com> > > --- > > Changes in v4: > > * Roundup the ring parameter value to a power of 2 > > * Skip the max value check for parameters > > * Use extack to log errors > > --- > > Changes in v3: > > * pre-allocate buffers before changing the queue sizes > > * rebased to latest net-next > > --- > > Changes in v2: > > * Removed unnecessary validations in mana_set_ringparam() > > * Fixed codespell error > > * Improved error message to indicate issue with the parameter > > --- > > drivers/net/ethernet/microsoft/mana/mana_en.c | 24 +++--- > > .../ethernet/microsoft/mana/mana_ethtool.c | 74 +++++++++++++++++++ > > include/net/mana/mana.h | 23 +++++- > > 3 files changed, 108 insertions(+), 13 deletions(-) > > > > diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c b/drivers/net/ethernet/microsoft/mana/mana_en.c > > index d2f07e179e86..4e3ade5926bc 100644 > > --- a/drivers/net/ethernet/microsoft/mana/mana_en.c > > +++ b/drivers/net/ethernet/microsoft/mana/mana_en.c > > @@ -511,7 +511,7 @@ static u16 mana_select_queue(struct net_device *ndev, struct sk_buff *skb, > > } > > > > /* Release pre-allocated RX buffers */ > > -static void mana_pre_dealloc_rxbufs(struct mana_port_context *mpc) > > +void mana_pre_dealloc_rxbufs(struct mana_port_context *mpc) > > { > > struct device *dev; > > int i; > > @@ -604,7 +604,7 @@ static void mana_get_rxbuf_cfg(int mtu, u32 *datasize, u32 *alloc_size, > > *datasize = mtu + ETH_HLEN; > > } > > > > -static int mana_pre_alloc_rxbufs(struct mana_port_context *mpc, int new_mtu) > > +int mana_pre_alloc_rxbufs(struct mana_port_context *mpc, int new_mtu) > > { > > struct device *dev; > > struct page *page; > > @@ -618,7 +618,7 @@ static int mana_pre_alloc_rxbufs(struct mana_port_context *mpc, int new_mtu) > > > > dev = mpc->ac->gdma_dev->gdma_context->dev; > > > > - num_rxb = mpc->num_queues * RX_BUFFERS_PER_QUEUE; > > + num_rxb = mpc->num_queues * mpc->rx_queue_size; > > > > WARN(mpc->rxbufs_pre, "mana rxbufs_pre exists\n"); > > mpc->rxbufs_pre = kmalloc_array(num_rxb, sizeof(void *), GFP_KERNEL); > > @@ -1899,14 +1899,15 @@ static int mana_create_txq(struct mana_port_context *apc, > > return -ENOMEM; > > > > /* The minimum size of the WQE is 32 bytes, hence > > - * MAX_SEND_BUFFERS_PER_QUEUE represents the maximum number of WQEs > > + * apc->tx_queue_size represents the maximum number of WQEs > > * the SQ can store. This value is then used to size other queues > > * to prevent overflow. > > + * Also note that the txq_size is always going to be MANA_PAGE_ALIGNED, > > + * as tx_queue_size is always a power of 2. > > */ > > MANA_PAGE_ALIGNED aligned means aligned by 0x1000. tx_queue_size being > 'power of 2' * 32 is not a sufficient condition for it to be aligned to > 0x1000. We possibly can explain more. > > > > - txq_size = MAX_SEND_BUFFERS_PER_QUEUE * 32; > > - BUILD_BUG_ON(!MANA_PAGE_ALIGNED(txq_size)); > > + txq_size = apc->tx_queue_size * 32; > > > > - cq_size = MAX_SEND_BUFFERS_PER_QUEUE * COMP_ENTRY_SIZE; > > + cq_size = apc->tx_queue_size * COMP_ENTRY_SIZE; > > cq_size = MANA_PAGE_ALIGN(cq_size); > > COMP_ENTRY_SIZE is 64, that means cq_size is double of txq_size. > If we are certain that txq_size is always aligned to MANA_PAGE, > that means cq_size is already aligned to MANA_PAGE as well. > > - Saurabh Thanks Saurabh. I'll incorporate these in the next version
diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c b/drivers/net/ethernet/microsoft/mana/mana_en.c index d2f07e179e86..4e3ade5926bc 100644 --- a/drivers/net/ethernet/microsoft/mana/mana_en.c +++ b/drivers/net/ethernet/microsoft/mana/mana_en.c @@ -511,7 +511,7 @@ static u16 mana_select_queue(struct net_device *ndev, struct sk_buff *skb, } /* Release pre-allocated RX buffers */ -static void mana_pre_dealloc_rxbufs(struct mana_port_context *mpc) +void mana_pre_dealloc_rxbufs(struct mana_port_context *mpc) { struct device *dev; int i; @@ -604,7 +604,7 @@ static void mana_get_rxbuf_cfg(int mtu, u32 *datasize, u32 *alloc_size, *datasize = mtu + ETH_HLEN; } -static int mana_pre_alloc_rxbufs(struct mana_port_context *mpc, int new_mtu) +int mana_pre_alloc_rxbufs(struct mana_port_context *mpc, int new_mtu) { struct device *dev; struct page *page; @@ -618,7 +618,7 @@ static int mana_pre_alloc_rxbufs(struct mana_port_context *mpc, int new_mtu) dev = mpc->ac->gdma_dev->gdma_context->dev; - num_rxb = mpc->num_queues * RX_BUFFERS_PER_QUEUE; + num_rxb = mpc->num_queues * mpc->rx_queue_size; WARN(mpc->rxbufs_pre, "mana rxbufs_pre exists\n"); mpc->rxbufs_pre = kmalloc_array(num_rxb, sizeof(void *), GFP_KERNEL); @@ -1899,14 +1899,15 @@ static int mana_create_txq(struct mana_port_context *apc, return -ENOMEM; /* The minimum size of the WQE is 32 bytes, hence - * MAX_SEND_BUFFERS_PER_QUEUE represents the maximum number of WQEs + * apc->tx_queue_size represents the maximum number of WQEs * the SQ can store. This value is then used to size other queues * to prevent overflow. + * Also note that the txq_size is always going to be MANA_PAGE_ALIGNED, + * as tx_queue_size is always a power of 2. */ - txq_size = MAX_SEND_BUFFERS_PER_QUEUE * 32; - BUILD_BUG_ON(!MANA_PAGE_ALIGNED(txq_size)); + txq_size = apc->tx_queue_size * 32; - cq_size = MAX_SEND_BUFFERS_PER_QUEUE * COMP_ENTRY_SIZE; + cq_size = apc->tx_queue_size * COMP_ENTRY_SIZE; cq_size = MANA_PAGE_ALIGN(cq_size); gc = gd->gdma_context; @@ -2145,10 +2146,11 @@ static int mana_push_wqe(struct mana_rxq *rxq) static int mana_create_page_pool(struct mana_rxq *rxq, struct gdma_context *gc) { + struct mana_port_context *mpc = netdev_priv(rxq->ndev); struct page_pool_params pprm = {}; int ret; - pprm.pool_size = RX_BUFFERS_PER_QUEUE; + pprm.pool_size = mpc->rx_queue_size; pprm.nid = gc->numa_node; pprm.napi = &rxq->rx_cq.napi; pprm.netdev = rxq->ndev; @@ -2180,13 +2182,13 @@ static struct mana_rxq *mana_create_rxq(struct mana_port_context *apc, gc = gd->gdma_context; - rxq = kzalloc(struct_size(rxq, rx_oobs, RX_BUFFERS_PER_QUEUE), + rxq = kzalloc(struct_size(rxq, rx_oobs, apc->rx_queue_size), GFP_KERNEL); if (!rxq) return NULL; rxq->ndev = ndev; - rxq->num_rx_buf = RX_BUFFERS_PER_QUEUE; + rxq->num_rx_buf = apc->rx_queue_size; rxq->rxq_idx = rxq_idx; rxq->rxobj = INVALID_MANA_HANDLE; @@ -2734,6 +2736,8 @@ static int mana_probe_port(struct mana_context *ac, int port_idx, apc->ndev = ndev; apc->max_queues = gc->max_num_queues; apc->num_queues = gc->max_num_queues; + apc->tx_queue_size = DEF_TX_BUFFERS_PER_QUEUE; + apc->rx_queue_size = DEF_RX_BUFFERS_PER_QUEUE; apc->port_handle = INVALID_MANA_HANDLE; apc->pf_filter_handle = INVALID_MANA_HANDLE; apc->port_idx = port_idx; diff --git a/drivers/net/ethernet/microsoft/mana/mana_ethtool.c b/drivers/net/ethernet/microsoft/mana/mana_ethtool.c index 146d5db1792f..d6a35fbda447 100644 --- a/drivers/net/ethernet/microsoft/mana/mana_ethtool.c +++ b/drivers/net/ethernet/microsoft/mana/mana_ethtool.c @@ -369,6 +369,78 @@ static int mana_set_channels(struct net_device *ndev, return err; } +static void mana_get_ringparam(struct net_device *ndev, + struct ethtool_ringparam *ring, + struct kernel_ethtool_ringparam *kernel_ring, + struct netlink_ext_ack *extack) +{ + struct mana_port_context *apc = netdev_priv(ndev); + + ring->rx_pending = apc->rx_queue_size; + ring->tx_pending = apc->tx_queue_size; + ring->rx_max_pending = MAX_RX_BUFFERS_PER_QUEUE; + ring->tx_max_pending = MAX_TX_BUFFERS_PER_QUEUE; +} + +static int mana_set_ringparam(struct net_device *ndev, + struct ethtool_ringparam *ring, + struct kernel_ethtool_ringparam *kernel_ring, + struct netlink_ext_ack *extack) +{ + struct mana_port_context *apc = netdev_priv(ndev); + u32 new_tx, new_rx; + u32 old_tx, old_rx; + int err; + + old_tx = apc->tx_queue_size; + old_rx = apc->rx_queue_size; + + if (ring->tx_pending < MIN_TX_BUFFERS_PER_QUEUE) { + NL_SET_ERR_MSG_FMT(extack, "tx:%d less than the min:%d", ring->tx_pending, + MIN_TX_BUFFERS_PER_QUEUE); + return -EINVAL; + } + + if (ring->rx_pending < MIN_RX_BUFFERS_PER_QUEUE) { + NL_SET_ERR_MSG_FMT(extack, "rx:%d less than the min:%d", ring->rx_pending, + MIN_RX_BUFFERS_PER_QUEUE); + return -EINVAL; + } + + new_rx = roundup_pow_of_two(ring->rx_pending); + new_tx = roundup_pow_of_two(ring->tx_pending); + netdev_info(ndev, "Using nearest power of 2 values for Txq:%d Rxq:%d\n", + new_tx, new_rx); + + /* pre-allocating new buffers to prevent failures in mana_attach() later */ + apc->rx_queue_size = new_rx; + err = mana_pre_alloc_rxbufs(apc, ndev->mtu); + apc->rx_queue_size = old_rx; + if (err) { + netdev_err(ndev, "Insufficient memory for new allocations\n"); + return err; + } + + err = mana_detach(ndev, false); + if (err) { + netdev_err(ndev, "mana_detach failed: %d\n", err); + goto out; + } + + apc->tx_queue_size = new_tx; + apc->rx_queue_size = new_rx; + + err = mana_attach(ndev); + if (err) { + netdev_err(ndev, "mana_attach failed: %d\n", err); + apc->tx_queue_size = old_tx; + apc->rx_queue_size = old_rx; + } +out: + mana_pre_dealloc_rxbufs(apc); + return err; +} + const struct ethtool_ops mana_ethtool_ops = { .get_ethtool_stats = mana_get_ethtool_stats, .get_sset_count = mana_get_sset_count, @@ -380,4 +452,6 @@ const struct ethtool_ops mana_ethtool_ops = { .set_rxfh = mana_set_rxfh, .get_channels = mana_get_channels, .set_channels = mana_set_channels, + .get_ringparam = mana_get_ringparam, + .set_ringparam = mana_set_ringparam, }; diff --git a/include/net/mana/mana.h b/include/net/mana/mana.h index 6439fd8b437b..80a1e53471a6 100644 --- a/include/net/mana/mana.h +++ b/include/net/mana/mana.h @@ -38,9 +38,21 @@ enum TRI_STATE { #define COMP_ENTRY_SIZE 64 -#define RX_BUFFERS_PER_QUEUE 512 +/* This Max value for RX buffers is derived from __alloc_page()'s max page + * allocation calculation. It allows maximum 2^(MAX_ORDER -1) pages. RX buffer + * size beyond this value gets rejected by __alloc_page() call. + */ +#define MAX_RX_BUFFERS_PER_QUEUE 8192 +#define DEF_RX_BUFFERS_PER_QUEUE 512 +#define MIN_RX_BUFFERS_PER_QUEUE 128 -#define MAX_SEND_BUFFERS_PER_QUEUE 256 +/* This max value for TX buffers is derived as the maximum allocatable + * pages supported on host per guest through testing. TX buffer size beyond + * this value is rejected by the hardware. + */ +#define MAX_TX_BUFFERS_PER_QUEUE 16384 +#define DEF_TX_BUFFERS_PER_QUEUE 256 +#define MIN_TX_BUFFERS_PER_QUEUE 128 #define EQ_SIZE (8 * MANA_PAGE_SIZE) @@ -285,7 +297,7 @@ struct mana_recv_buf_oob { void *buf_va; bool from_pool; /* allocated from a page pool */ - /* SGL of the buffer going to be sent has part of the work request. */ + /* SGL of the buffer going to be sent as part of the work request. */ u32 num_sge; struct gdma_sge sgl[MAX_RX_WQE_SGL_ENTRIES]; @@ -437,6 +449,9 @@ struct mana_port_context { unsigned int max_queues; unsigned int num_queues; + unsigned int rx_queue_size; + unsigned int tx_queue_size; + mana_handle_t port_handle; mana_handle_t pf_filter_handle; @@ -472,6 +487,8 @@ struct bpf_prog *mana_xdp_get(struct mana_port_context *apc); void mana_chn_setxdp(struct mana_port_context *apc, struct bpf_prog *prog); int mana_bpf(struct net_device *ndev, struct netdev_bpf *bpf); void mana_query_gf_stats(struct mana_port_context *apc); +int mana_pre_alloc_rxbufs(struct mana_port_context *apc, int mtu); +void mana_pre_dealloc_rxbufs(struct mana_port_context *apc); extern const struct ethtool_ops mana_ethtool_ops;