* [Intel-wired-lan] [PATCH net 1/5] idpf: add support for Tx refillqs in flow scheduling mode
2025-06-25 16:11 [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool Joshua Hay
@ 2025-06-25 16:11 ` Joshua Hay
2025-06-27 13:13 ` Alexander Lobakin
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 2/5] idpf: improve when to set RE bit logic Joshua Hay
` (4 subsequent siblings)
5 siblings, 1 reply; 16+ messages in thread
From: Joshua Hay @ 2025-06-25 16:11 UTC (permalink / raw)
To: intel-wired-lan; +Cc: netdev, Joshua Hay, Madhu Chittim
This is the start of a 5 patch series intended to fix a stability issue
in the flow scheduling Tx send/clean path that results in a Tx timeout.
In certain production environments, it is possible for completion tags
to collide, meaning N packets with the same completion tag are in flight
at the same time. In this environment, any given Tx queue is effectively
used to send both slower traffic and higher throughput traffic
simultaneously. This is the result of a customer's specific
configuration in the device pipeline, the details of which Intel cannot
provide. This configuration results in a small number of out-of-order
completions, i.e., a small number of packets in flight. The existing
guardrails in the driver only protect against a large number of packets
in flight. The slower flow completions are delayed which causes the
out-of-order completions. Meanwhile, the fast flow exhausts the pool of
unique tags and starts reusing tags. The next packet in the fast flow
uses the same tag for a packet that is still in flight from the slower
flow. The driver has no idea which packet it should clean when it
processes the completion with that tag, but it will for the packet on
the buffer ring before the hash table. If the slower flow packet
completion is processed first, it will end up cleaning the fast flow
packet on the ring prematurely. This leaves the descriptor ring in a bad
state resulting in a Tx timeout.
This series refactors the Tx buffer management by replacing the stashing
mechanisms and the tag generation with a large pool/array of unique
tags. The completion tags are now simply used to index into the pool of
Tx buffers. This implicitly prevents any tag from being reused while
it's in flight.
First, we need a new mechanism for the send path to know what tag to use
next. The driver will allocate and initialize a refillq for each TxQ
with all of the possible free tag values. During send, the driver grabs
the next free tag from the refillq from next_to_clean. While cleaning
the packet, the clean routine posts the tag back to the refillq's
next_to_use to indicate that it is now free to use.
This mechanism works exactly the same way as the existing Rx refill
queues, which post the cleaned buffer IDs back to the buffer queue to be
reposted to HW. Since we're using the refillqs for both Rx and Tx now,
genercize some of the existing refillq support.
Note: the refillqs will not be used yet. This is only demonstrating how
they will be used to pass free tags back to the send path.
Signed-off-by: Joshua Hay <joshua.a.hay@intel.com>
Reviewed-by: Madhu Chittim <madhu.chittim@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf_txrx.c | 91 +++++++++++++++++++--
drivers/net/ethernet/intel/idpf/idpf_txrx.h | 8 +-
2 files changed, 89 insertions(+), 10 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.c b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
index 5cf440e09d0a..6a16b80a8ac2 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
@@ -139,6 +139,9 @@ static void idpf_tx_desc_rel(struct idpf_tx_queue *txq)
if (!txq->desc_ring)
return;
+ if (txq->refillq)
+ kfree(txq->refillq->ring);
+
dmam_free_coherent(txq->dev, txq->size, txq->desc_ring, txq->dma);
txq->desc_ring = NULL;
txq->next_to_use = 0;
@@ -267,6 +270,31 @@ static int idpf_tx_desc_alloc(const struct idpf_vport *vport,
tx_q->next_to_clean = 0;
idpf_queue_set(GEN_CHK, tx_q);
+ if (idpf_queue_has(FLOW_SCH_EN, tx_q)) {
+ struct idpf_sw_queue *refillq = tx_q->refillq;
+
+ refillq->desc_count = tx_q->desc_count;
+
+ refillq->ring = kcalloc(refillq->desc_count, sizeof(u32),
+ GFP_KERNEL);
+ if (!refillq->ring) {
+ err = -ENOMEM;
+ goto err_alloc;
+ }
+
+ for (u32 i = 0; i < refillq->desc_count; i++)
+ refillq->ring[i] =
+ FIELD_PREP(IDPF_RFL_BI_BUFID_M, i) |
+ FIELD_PREP(IDPF_RFL_BI_GEN_M,
+ idpf_queue_has(GEN_CHK, refillq));
+
+ /*
+ * Go ahead and flip the GEN bit since this counts as filling
+ * up the ring, i.e. we already ring wrapped.
+ */
+ idpf_queue_change(GEN_CHK, refillq);
+ }
+
return 0;
err_alloc:
@@ -603,18 +631,18 @@ static int idpf_rx_hdr_buf_alloc_all(struct idpf_buf_queue *bufq)
}
/**
- * idpf_rx_post_buf_refill - Post buffer id to refill queue
+ * idpf_post_buf_refill - Post buffer id to refill queue
* @refillq: refill queue to post to
* @buf_id: buffer id to post
*/
-static void idpf_rx_post_buf_refill(struct idpf_sw_queue *refillq, u16 buf_id)
+static void idpf_post_buf_refill(struct idpf_sw_queue *refillq, u16 buf_id)
{
u32 nta = refillq->next_to_use;
/* store the buffer ID and the SW maintained GEN bit to the refillq */
refillq->ring[nta] =
- FIELD_PREP(IDPF_RX_BI_BUFID_M, buf_id) |
- FIELD_PREP(IDPF_RX_BI_GEN_M,
+ FIELD_PREP(IDPF_RFL_BI_BUFID_M, buf_id) |
+ FIELD_PREP(IDPF_RFL_BI_GEN_M,
idpf_queue_has(GEN_CHK, refillq));
if (unlikely(++nta == refillq->desc_count)) {
@@ -995,6 +1023,11 @@ static void idpf_txq_group_rel(struct idpf_vport *vport)
struct idpf_txq_group *txq_grp = &vport->txq_grps[i];
for (j = 0; j < txq_grp->num_txq; j++) {
+ if (flow_sch_en) {
+ kfree(txq_grp->txqs[j]->refillq);
+ txq_grp->txqs[j]->refillq = NULL;
+ }
+
kfree(txq_grp->txqs[j]);
txq_grp->txqs[j] = NULL;
}
@@ -1414,6 +1447,13 @@ static int idpf_txq_group_alloc(struct idpf_vport *vport, u16 num_txq)
}
idpf_queue_set(FLOW_SCH_EN, q);
+
+ q->refillq = kzalloc(sizeof(*q->refillq), GFP_KERNEL);
+ if (!q->refillq)
+ goto err_alloc;
+
+ idpf_queue_set(GEN_CHK, q->refillq);
+ idpf_queue_set(RFL_GEN_CHK, q->refillq);
}
if (!split)
@@ -2005,6 +2045,8 @@ static void idpf_tx_handle_rs_completion(struct idpf_tx_queue *txq,
compl_tag = le16_to_cpu(desc->q_head_compl_tag.compl_tag);
+ idpf_post_buf_refill(txq->refillq, compl_tag);
+
/* If we didn't clean anything on the ring, this packet must be
* in the hash table. Go clean it there.
*/
@@ -2364,6 +2406,37 @@ static unsigned int idpf_tx_splitq_bump_ntu(struct idpf_tx_queue *txq, u16 ntu)
return ntu;
}
+/**
+ * idpf_tx_get_free_buf_id - get a free buffer ID from the refill queue
+ * @refillq: refill queue to get buffer ID from
+ * @buf_id: return buffer ID
+ *
+ * Return: true if a buffer ID was found, false if not
+ */
+static bool idpf_tx_get_free_buf_id(struct idpf_sw_queue *refillq,
+ u16 *buf_id)
+{
+ u16 ntc = refillq->next_to_clean;
+ u32 refill_desc;
+
+ refill_desc = refillq->ring[ntc];
+
+ if (idpf_queue_has(RFL_GEN_CHK, refillq) !=
+ !!(refill_desc & IDPF_RFL_BI_GEN_M))
+ return false;
+
+ *buf_id = FIELD_GET(IDPF_RFL_BI_BUFID_M, refill_desc);
+
+ if (unlikely(++ntc == refillq->desc_count)) {
+ idpf_queue_change(RFL_GEN_CHK, refillq);
+ ntc = 0;
+ }
+
+ refillq->next_to_clean = ntc;
+
+ return true;
+}
+
/**
* idpf_tx_splitq_map - Build the Tx flex descriptor
* @tx_q: queue to send buffer on
@@ -2912,6 +2985,10 @@ static netdev_tx_t idpf_tx_splitq_frame(struct sk_buff *skb,
}
if (idpf_queue_has(FLOW_SCH_EN, tx_q)) {
+ if (unlikely(!idpf_tx_get_free_buf_id(tx_q->refillq,
+ &tx_params.compl_tag)))
+ return idpf_tx_drop_skb(tx_q, skb);
+
tx_params.dtype = IDPF_TX_DESC_DTYPE_FLEX_FLOW_SCHE;
tx_params.eop_cmd = IDPF_TXD_FLEX_FLOW_CMD_EOP;
/* Set the RE bit to catch any packets that may have not been
@@ -3464,7 +3541,7 @@ static int idpf_rx_splitq_clean(struct idpf_rx_queue *rxq, int budget)
skip_data:
rx_buf->page = NULL;
- idpf_rx_post_buf_refill(refillq, buf_id);
+ idpf_post_buf_refill(refillq, buf_id);
IDPF_RX_BUMP_NTC(rxq, ntc);
/* skip if it is non EOP desc */
@@ -3572,10 +3649,10 @@ static void idpf_rx_clean_refillq(struct idpf_buf_queue *bufq,
bool failure;
if (idpf_queue_has(RFL_GEN_CHK, refillq) !=
- !!(refill_desc & IDPF_RX_BI_GEN_M))
+ !!(refill_desc & IDPF_RFL_BI_GEN_M))
break;
- buf_id = FIELD_GET(IDPF_RX_BI_BUFID_M, refill_desc);
+ buf_id = FIELD_GET(IDPF_RFL_BI_BUFID_M, refill_desc);
failure = idpf_rx_update_bufq_desc(bufq, buf_id, buf_desc);
if (failure)
break;
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.h b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
index 36a0f828a6f8..6924bee6ff5b 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.h
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
@@ -107,8 +107,8 @@ do { \
*/
#define IDPF_TX_SPLITQ_RE_MIN_GAP 64
-#define IDPF_RX_BI_GEN_M BIT(16)
-#define IDPF_RX_BI_BUFID_M GENMASK(15, 0)
+#define IDPF_RFL_BI_GEN_M BIT(16)
+#define IDPF_RFL_BI_BUFID_M GENMASK(15, 0)
#define IDPF_RXD_EOF_SPLITQ VIRTCHNL2_RX_FLEX_DESC_ADV_STATUS0_EOF_M
#define IDPF_RXD_EOF_SINGLEQ VIRTCHNL2_RX_BASE_DESC_STATUS_EOF_M
@@ -621,6 +621,7 @@ libeth_cacheline_set_assert(struct idpf_rx_queue, 64,
* @cleaned_pkts: Number of packets cleaned for the above said case
* @tx_max_bufs: Max buffers that can be transmitted with scatter-gather
* @stash: Tx buffer stash for Flow-based scheduling mode
+ * @refillq: Pointer to refill queue
* @compl_tag_bufid_m: Completion tag buffer id mask
* @compl_tag_cur_gen: Used to keep track of current completion tag generation
* @compl_tag_gen_max: To determine when compl_tag_cur_gen should be reset
@@ -670,6 +671,7 @@ struct idpf_tx_queue {
u16 tx_max_bufs;
struct idpf_txq_stash *stash;
+ struct idpf_sw_queue *refillq;
u16 compl_tag_bufid_m;
u16 compl_tag_cur_gen;
@@ -691,7 +693,7 @@ struct idpf_tx_queue {
__cacheline_group_end_aligned(cold);
};
libeth_cacheline_set_assert(struct idpf_tx_queue, 64,
- 112 + sizeof(struct u64_stats_sync),
+ 120 + sizeof(struct u64_stats_sync),
24);
/**
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread* Re: [Intel-wired-lan] [PATCH net 1/5] idpf: add support for Tx refillqs in flow scheduling mode
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 1/5] idpf: add support for Tx refillqs in flow scheduling mode Joshua Hay
@ 2025-06-27 13:13 ` Alexander Lobakin
0 siblings, 0 replies; 16+ messages in thread
From: Alexander Lobakin @ 2025-06-27 13:13 UTC (permalink / raw)
To: Joshua Hay; +Cc: intel-wired-lan, netdev, Madhu Chittim
From: Joshua Hay <joshua.a.hay@intel.com>
Date: Wed, 25 Jun 2025 09:11:52 -0700
> This is the start of a 5 patch series intended to fix a stability issue
> in the flow scheduling Tx send/clean path that results in a Tx timeout.
No need to mention "series", "start", "patch" in commit messages.
>
> In certain production environments, it is possible for completion tags
> to collide, meaning N packets with the same completion tag are in flight
> at the same time. In this environment, any given Tx queue is effectively
> used to send both slower traffic and higher throughput traffic
> simultaneously. This is the result of a customer's specific
> configuration in the device pipeline, the details of which Intel cannot
> provide. This configuration results in a small number of out-of-order
> completions, i.e., a small number of packets in flight. The existing
> guardrails in the driver only protect against a large number of packets
> in flight. The slower flow completions are delayed which causes the
> out-of-order completions. Meanwhile, the fast flow exhausts the pool of
> unique tags and starts reusing tags. The next packet in the fast flow
> uses the same tag for a packet that is still in flight from the slower
> flow. The driver has no idea which packet it should clean when it
> processes the completion with that tag, but it will for the packet on
> the buffer ring before the hash table. If the slower flow packet
> completion is processed first, it will end up cleaning the fast flow
> packet on the ring prematurely. This leaves the descriptor ring in a bad
> state resulting in a Tx timeout.
>
> This series refactors the Tx buffer management by replacing the stashing
Same.
> mechanisms and the tag generation with a large pool/array of unique
> tags. The completion tags are now simply used to index into the pool of
> Tx buffers. This implicitly prevents any tag from being reused while
> it's in flight.
>
> First, we need a new mechanism for the send path to know what tag to use
> next. The driver will allocate and initialize a refillq for each TxQ
> with all of the possible free tag values. During send, the driver grabs
> the next free tag from the refillq from next_to_clean. While cleaning
> the packet, the clean routine posts the tag back to the refillq's
> next_to_use to indicate that it is now free to use.
>
> This mechanism works exactly the same way as the existing Rx refill
> queues, which post the cleaned buffer IDs back to the buffer queue to be
> reposted to HW. Since we're using the refillqs for both Rx and Tx now,
> genercize some of the existing refillq support.
>
> Note: the refillqs will not be used yet. This is only demonstrating how
> they will be used to pass free tags back to the send path.
[...]
> @@ -267,6 +270,31 @@ static int idpf_tx_desc_alloc(const struct idpf_vport *vport,
> tx_q->next_to_clean = 0;
> idpf_queue_set(GEN_CHK, tx_q);
>
> + if (idpf_queue_has(FLOW_SCH_EN, tx_q)) {
if (!idpf_queue_has(FLOW_SCH_EN, tx_q))
return 0;
> + struct idpf_sw_queue *refillq = tx_q->refillq;
> +
> + refillq->desc_count = tx_q->desc_count;
> +
> + refillq->ring = kcalloc(refillq->desc_count, sizeof(u32),
> + GFP_KERNEL);
> + if (!refillq->ring) {
> + err = -ENOMEM;
> + goto err_alloc;
> + }
> +
> + for (u32 i = 0; i < refillq->desc_count; i++)
> + refillq->ring[i] =
> + FIELD_PREP(IDPF_RFL_BI_BUFID_M, i) |
> + FIELD_PREP(IDPF_RFL_BI_GEN_M,
> + idpf_queue_has(GEN_CHK, refillq));
> +
> + /*
> + * Go ahead and flip the GEN bit since this counts as filling
> + * up the ring, i.e. we already ring wrapped.
> + */
> + idpf_queue_change(GEN_CHK, refillq);
> + }
> +
> return 0;
>
> err_alloc:
Thanks,
Olek
^ permalink raw reply [flat|nested] 16+ messages in thread
* [Intel-wired-lan] [PATCH net 2/5] idpf: improve when to set RE bit logic
2025-06-25 16:11 [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool Joshua Hay
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 1/5] idpf: add support for Tx refillqs in flow scheduling mode Joshua Hay
@ 2025-06-25 16:11 ` Joshua Hay
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 3/5] idpf: replace flow scheduling buffer ring with buffer pool Joshua Hay
` (3 subsequent siblings)
5 siblings, 0 replies; 16+ messages in thread
From: Joshua Hay @ 2025-06-25 16:11 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, Joshua Hay, Luigi Rizzo, Brian Vazquez, Madhu Chittim
Track the gap between next_to_use and the last RE index. Set RE again
if the gap is large enough to ensure RE bit is set frequently. This is
critical before removing the stashing mechanisms because the
opportunistic descriptor ring cleaning from the out-of-order completions
will go away. Previously the descriptors would be "cleaned" by both the
descriptor (RE) completion and the out-of-order completions. Without the
latter, we must ensure the RE bit is set more frequently. Otherwise,
it's theoretically possible for the descriptor ring next_to_clean to
never advance. The previous implementation was dependent on the start
of a packet falling on a 64th index in the descriptor ring, which is not
guaranteed with large packets.
Signed-off-by: Luigi Rizzo <lrizzo@google.com>
Signed-off-by: Brian Vazquez <brianvv@google.com>
Signed-off-by: Joshua Hay <joshua.a.hay@intel.com>
Reviewed-by: Madhu Chittim <madhu.chittim@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf_txrx.c | 20 +++++++++++++++++++-
drivers/net/ethernet/intel/idpf/idpf_txrx.h | 6 ++++--
2 files changed, 23 insertions(+), 3 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.c b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
index 6a16b80a8ac2..cdecf558d7ec 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
@@ -293,6 +293,8 @@ static int idpf_tx_desc_alloc(const struct idpf_vport *vport,
* up the ring, i.e. we already ring wrapped.
*/
idpf_queue_change(GEN_CHK, refillq);
+
+ tx_q->last_re = tx_q->desc_count - IDPF_TX_SPLITQ_RE_MIN_GAP;
}
return 0;
@@ -2913,6 +2915,21 @@ static void idpf_tx_set_tstamp_desc(union idpf_flex_tx_ctx_desc *ctx_desc,
{ }
#endif /* CONFIG_PTP_1588_CLOCK */
+/**
+ * idpf_tx_splitq_need_re - check whether RE bit needs to be set
+ * @tx_q: pointer to Tx queue
+ *
+ * Return: true if RE bit needs to be set, false otherwise
+ */
+static bool idpf_tx_splitq_need_re(struct idpf_tx_queue *tx_q)
+{
+ int gap = tx_q->next_to_use - tx_q->last_re;
+
+ gap += (gap < 0) ? tx_q->desc_count : 0;
+
+ return gap >= IDPF_TX_SPLITQ_RE_MIN_GAP;
+}
+
/**
* idpf_tx_splitq_frame - Sends buffer on Tx ring using flex descriptors
* @skb: send buffer
@@ -2996,9 +3013,10 @@ static netdev_tx_t idpf_tx_splitq_frame(struct sk_buff *skb,
* MIN_RING size to ensure it will be set at least once each
* time around the ring.
*/
- if (!(tx_q->next_to_use % IDPF_TX_SPLITQ_RE_MIN_GAP)) {
+ if (idpf_tx_splitq_need_re(tx_q)) {
tx_params.eop_cmd |= IDPF_TXD_FLEX_FLOW_CMD_RE;
tx_q->txq_grp->num_completions_pending++;
+ tx_q->last_re = tx_q->next_to_use;
}
if (skb->ip_summed == CHECKSUM_PARTIAL)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.h b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
index 6924bee6ff5b..65acad4c929d 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.h
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
@@ -609,6 +609,8 @@ libeth_cacheline_set_assert(struct idpf_rx_queue, 64,
* @netdev: &net_device corresponding to this queue
* @next_to_use: Next descriptor to use
* @next_to_clean: Next descriptor to clean
+ * @last_re: last descriptor index that RE bit was set
+ * @tx_max_bufs: Max buffers that can be transmitted with scatter-gather
* @cleaned_bytes: Splitq only, TXQ only: When a TX completion is received on
* the TX completion queue, it can be for any TXQ associated
* with that completion queue. This means we can clean up to
@@ -619,7 +621,6 @@ libeth_cacheline_set_assert(struct idpf_rx_queue, 64,
* only once at the end of the cleaning routine.
* @clean_budget: singleq only, queue cleaning budget
* @cleaned_pkts: Number of packets cleaned for the above said case
- * @tx_max_bufs: Max buffers that can be transmitted with scatter-gather
* @stash: Tx buffer stash for Flow-based scheduling mode
* @refillq: Pointer to refill queue
* @compl_tag_bufid_m: Completion tag buffer id mask
@@ -662,6 +663,8 @@ struct idpf_tx_queue {
__cacheline_group_begin_aligned(read_write);
u16 next_to_use;
u16 next_to_clean;
+ u16 last_re;
+ u16 tx_max_bufs;
union {
u32 cleaned_bytes;
@@ -669,7 +672,6 @@ struct idpf_tx_queue {
};
u16 cleaned_pkts;
- u16 tx_max_bufs;
struct idpf_txq_stash *stash;
struct idpf_sw_queue *refillq;
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread* [Intel-wired-lan] [PATCH net 3/5] idpf: replace flow scheduling buffer ring with buffer pool
2025-06-25 16:11 [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool Joshua Hay
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 1/5] idpf: add support for Tx refillqs in flow scheduling mode Joshua Hay
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 2/5] idpf: improve when to set RE bit logic Joshua Hay
@ 2025-06-25 16:11 ` Joshua Hay
2025-06-27 13:10 ` Alexander Lobakin
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 4/5] idpf: stop Tx if there are insufficient buffer resources Joshua Hay
` (2 subsequent siblings)
5 siblings, 1 reply; 16+ messages in thread
From: Joshua Hay @ 2025-06-25 16:11 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, Joshua Hay, Luigi Rizzo, Brian Vazquez, Madhu Chittim
Replace the TxQ buffer ring with one large pool/array of buffers (only
for flow scheduling). The completion tag passed to HW through the
descriptor is the index into this array. That same completion tag is
posted back to the driver in the completion descriptor, and used to
index into the array to quickly retrieve the buffer during cleaning. In
this way, the tags are treated as a fix sized resource. If all tags are
in use, no more packets can be sent on that particular queue (until some
are freed up). The tag pool size is 64K since the completion tag width
is 16 bits.
For each packet, the driver pulls a free tag from the refillq to get the
next free buffer index. When cleaning is complete, the tag is posted
back to the refillq. A multi-frag packet spans multiple buffers in the
driver, therefore it uses multiple buffer indexes/tags from the pool.
Each frag pulls from the refillq to get the next free buffer index.
These are tracked in a next_buf field that replaces the completion tag
field in the buffer struct. This chains the buffers together so that the
packet can be cleaned from the starting completion tag taken from the
completion descriptor, then from the next_buf field for each subsequent
buffer.
Descriptor completions only advance the descriptor ring index to "clean"
the descriptors. The packet completions only clean the buffers
associated with the given packet completion tag and do not update the
descriptor ring index.
When operating in queue based scheduling mode, the array still acts as a
ring and will only have TxQ descriptor count entries. The tx_bufs are
still associated 1:1 with the descriptor ring entries and we can use the
conventional indexing mechanisms.
Fixes: c2d548cad150 ("idpf: add TX splitq napi poll support")
Signed-off-by: Luigi Rizzo <lrizzo@google.com>
Signed-off-by: Brian Vazquez <brianvv@google.com>
Signed-off-by: Joshua Hay <joshua.a.hay@intel.com>
Reviewed-by: Madhu Chittim <madhu.chittim@intel.com>
---
.../ethernet/intel/idpf/idpf_singleq_txrx.c | 2 +-
drivers/net/ethernet/intel/idpf/idpf_txrx.c | 135 ++++++++++--------
drivers/net/ethernet/intel/idpf/idpf_txrx.h | 10 +-
3 files changed, 85 insertions(+), 62 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c b/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c
index 993c354aa27a..4c1d7235a073 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c
@@ -220,7 +220,7 @@ static void idpf_tx_singleq_map(struct idpf_tx_queue *tx_q,
unsigned int max_data = IDPF_TX_MAX_DESC_DATA_ALIGNED;
if (dma_mapping_error(tx_q->dev, dma))
- return idpf_tx_dma_map_error(tx_q, skb, first, i);
+ return idpf_tx_dma_map_unwind(tx_q, skb, first, i);
/* record length, and DMA address */
dma_unmap_len_set(tx_buf, len, size);
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.c b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
index cdecf558d7ec..25eea632a966 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
@@ -13,6 +13,7 @@ struct idpf_tx_stash {
struct libeth_sqe buf;
};
+#define idpf_tx_buf_next(buf) (*(u32 *)&(buf)->priv)
#define idpf_tx_buf_compl_tag(buf) (*(u32 *)&(buf)->priv)
LIBETH_SQE_CHECK_PRIV(u32);
@@ -91,7 +92,7 @@ static void idpf_tx_buf_rel_all(struct idpf_tx_queue *txq)
return;
/* Free all the Tx buffer sk_buffs */
- for (i = 0; i < txq->desc_count; i++)
+ for (i = 0; i < txq->buf_pool_size; i++)
libeth_tx_complete(&txq->tx_buf[i], &cp);
kfree(txq->tx_buf);
@@ -205,7 +206,11 @@ static int idpf_tx_buf_alloc_all(struct idpf_tx_queue *tx_q)
/* Allocate book keeping buffers only. Buffers to be supplied to HW
* are allocated by kernel network stack and received as part of skb
*/
- buf_size = sizeof(struct idpf_tx_buf) * tx_q->desc_count;
+ if (idpf_queue_has(FLOW_SCH_EN, tx_q))
+ tx_q->buf_pool_size = U16_MAX;
+ else
+ tx_q->buf_pool_size = tx_q->desc_count;
+ buf_size = sizeof(struct idpf_tx_buf) * tx_q->buf_pool_size;
tx_q->tx_buf = kzalloc(buf_size, GFP_KERNEL);
if (!tx_q->tx_buf)
return -ENOMEM;
@@ -273,7 +278,7 @@ static int idpf_tx_desc_alloc(const struct idpf_vport *vport,
if (idpf_queue_has(FLOW_SCH_EN, tx_q)) {
struct idpf_sw_queue *refillq = tx_q->refillq;
- refillq->desc_count = tx_q->desc_count;
+ refillq->desc_count = tx_q->buf_pool_size;
refillq->ring = kcalloc(refillq->desc_count, sizeof(u32),
GFP_KERNEL);
@@ -1870,6 +1875,12 @@ static bool idpf_tx_splitq_clean(struct idpf_tx_queue *tx_q, u16 end,
struct idpf_tx_buf *tx_buf;
bool clean_complete = true;
+ if (descs_only) {
+ /* Bump ring index to mark as cleaned. */
+ tx_q->next_to_clean = end;
+ return true;
+ }
+
tx_desc = &tx_q->flex_tx[ntc];
next_pending_desc = &tx_q->flex_tx[end];
tx_buf = &tx_q->tx_buf[ntc];
@@ -1936,22 +1947,20 @@ do { \
} while (0)
/**
- * idpf_tx_clean_buf_ring - clean flow scheduling TX queue buffers
+ * idpf_tx_clean_bufs - clean flow scheduling TX queue buffers
* @txq: queue to clean
- * @compl_tag: completion tag of packet to clean (from completion descriptor)
+ * @buf_id: packet's starting buffer ID, from completion descriptor
* @cleaned: pointer to stats struct to track cleaned packets/bytes
* @budget: Used to determine if we are in netpoll
*
- * Cleans all buffers associated with the input completion tag either from the
- * TX buffer ring or from the hash table if the buffers were previously
- * stashed. Returns the byte/segment count for the cleaned packet associated
- * this completion tag.
+ * Clean all buffers associated with the packet starting at buf_id. Returns the
+ * byte/segment count for the cleaned packet.
*/
-static bool idpf_tx_clean_buf_ring(struct idpf_tx_queue *txq, u16 compl_tag,
- struct libeth_sq_napi_stats *cleaned,
- int budget)
+static bool idpf_tx_clean_bufs(struct idpf_tx_queue *txq, u16 buf_id,
+ struct libeth_sq_napi_stats *cleaned,
+ int budget)
{
- u16 idx = compl_tag & txq->compl_tag_bufid_m;
+ u16 idx = buf_id & txq->compl_tag_bufid_m;
struct idpf_tx_buf *tx_buf = NULL;
struct libeth_cq_pp cp = {
.dev = txq->dev,
@@ -1960,10 +1969,29 @@ static bool idpf_tx_clean_buf_ring(struct idpf_tx_queue *txq, u16 compl_tag,
};
u16 ntc, orig_idx = idx;
+ tx_buf = &txq->tx_buf[buf_id];
+ if (tx_buf->type == LIBETH_SQE_SKB) {
+ if (skb_shinfo(tx_buf->skb)->tx_flags & SKBTX_IN_PROGRESS)
+ idpf_tx_read_tstamp(txq, tx_buf->skb);
+
+ libeth_tx_complete(tx_buf, &cp);
+ idpf_post_buf_refill(txq->refillq, buf_id);
+ }
+
+ while (idpf_tx_buf_next(tx_buf) != IDPF_TXBUF_NULL) {
+ u16 buf_id = idpf_tx_buf_next(tx_buf);
+
+ tx_buf = &txq->tx_buf[buf_id];
+ libeth_tx_complete(tx_buf, &cp);
+ idpf_post_buf_refill(txq->refillq, buf_id);
+ }
+
+ return true;
+
tx_buf = &txq->tx_buf[idx];
if (unlikely(tx_buf->type <= LIBETH_SQE_CTX ||
- idpf_tx_buf_compl_tag(tx_buf) != compl_tag))
+ idpf_tx_buf_compl_tag(tx_buf) != buf_id))
return false;
if (tx_buf->type == LIBETH_SQE_SKB) {
@@ -1975,7 +2003,7 @@ static bool idpf_tx_clean_buf_ring(struct idpf_tx_queue *txq, u16 compl_tag,
idpf_tx_clean_buf_ring_bump_ntc(txq, idx, tx_buf);
- while (idpf_tx_buf_compl_tag(tx_buf) == compl_tag) {
+ while (idpf_tx_buf_compl_tag(tx_buf) == buf_id) {
libeth_tx_complete(tx_buf, &cp);
idpf_tx_clean_buf_ring_bump_ntc(txq, idx, tx_buf);
}
@@ -2047,12 +2075,10 @@ static void idpf_tx_handle_rs_completion(struct idpf_tx_queue *txq,
compl_tag = le16_to_cpu(desc->q_head_compl_tag.compl_tag);
- idpf_post_buf_refill(txq->refillq, compl_tag);
-
/* If we didn't clean anything on the ring, this packet must be
* in the hash table. Go clean it there.
*/
- if (!idpf_tx_clean_buf_ring(txq, compl_tag, cleaned, budget))
+ if (!idpf_tx_clean_bufs(txq, compl_tag, cleaned, budget))
idpf_tx_clean_stashed_bufs(txq, compl_tag, cleaned, budget);
}
@@ -2341,14 +2367,14 @@ unsigned int idpf_tx_desc_count_required(struct idpf_tx_queue *txq,
}
/**
- * idpf_tx_dma_map_error - handle TX DMA map errors
+ * idpf_tx_dma_map_unwind - handle TX DMA map errors
* @txq: queue to send buffer on
* @skb: send buffer
* @first: original first buffer info buffer for packet
* @idx: starting point on ring to unwind
*/
-void idpf_tx_dma_map_error(struct idpf_tx_queue *txq, struct sk_buff *skb,
- struct idpf_tx_buf *first, u16 idx)
+void idpf_tx_dma_map_unwind(struct idpf_tx_queue *txq, struct sk_buff *skb,
+ struct idpf_tx_buf *first, u16 idx)
{
struct libeth_sq_napi_stats ss = { };
struct libeth_cq_pp cp = {
@@ -2460,6 +2486,7 @@ static void idpf_tx_splitq_map(struct idpf_tx_queue *tx_q,
struct netdev_queue *nq;
struct sk_buff *skb;
skb_frag_t *frag;
+ u16 next_buf_id;
u16 td_cmd = 0;
dma_addr_t dma;
@@ -2477,17 +2504,13 @@ static void idpf_tx_splitq_map(struct idpf_tx_queue *tx_q,
tx_buf = first;
first->nr_frags = 0;
- params->compl_tag =
- (tx_q->compl_tag_cur_gen << tx_q->compl_tag_gen_s) | i;
-
for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
unsigned int max_data = IDPF_TX_MAX_DESC_DATA_ALIGNED;
if (dma_mapping_error(tx_q->dev, dma))
- return idpf_tx_dma_map_error(tx_q, skb, first, i);
+ return idpf_tx_dma_map_unwind(tx_q, skb, first, i);
first->nr_frags++;
- idpf_tx_buf_compl_tag(tx_buf) = params->compl_tag;
tx_buf->type = LIBETH_SQE_FRAG;
/* record length, and DMA address */
@@ -2543,29 +2566,14 @@ static void idpf_tx_splitq_map(struct idpf_tx_queue *tx_q,
max_data);
if (unlikely(++i == tx_q->desc_count)) {
- tx_buf = tx_q->tx_buf;
tx_desc = &tx_q->flex_tx[0];
i = 0;
tx_q->compl_tag_cur_gen =
IDPF_TX_ADJ_COMPL_TAG_GEN(tx_q);
} else {
- tx_buf++;
tx_desc++;
}
- /* Since this packet has a buffer that is going to span
- * multiple descriptors, it's going to leave holes in
- * to the TX buffer ring. To ensure these holes do not
- * cause issues in the cleaning routines, we will clear
- * them of any stale data and assign them the same
- * completion tag as the current packet. Then when the
- * packet is being cleaned, the cleaning routines will
- * simply pass over these holes and finish cleaning the
- * rest of the packet.
- */
- tx_buf->type = LIBETH_SQE_EMPTY;
- idpf_tx_buf_compl_tag(tx_buf) = params->compl_tag;
-
/* Adjust the DMA offset and the remaining size of the
* fragment. On the first iteration of this loop,
* max_data will be >= 12K and <= 16K-1. On any
@@ -2590,15 +2598,24 @@ static void idpf_tx_splitq_map(struct idpf_tx_queue *tx_q,
idpf_tx_splitq_build_desc(tx_desc, params, td_cmd, size);
if (unlikely(++i == tx_q->desc_count)) {
- tx_buf = tx_q->tx_buf;
tx_desc = &tx_q->flex_tx[0];
i = 0;
tx_q->compl_tag_cur_gen = IDPF_TX_ADJ_COMPL_TAG_GEN(tx_q);
} else {
- tx_buf++;
tx_desc++;
}
+ if (idpf_queue_has(FLOW_SCH_EN, tx_q)) {
+ if (unlikely(!idpf_tx_get_free_buf_id(tx_q->refillq,
+ &next_buf_id)))
+ return idpf_tx_dma_map_unwind(tx_q, skb, first, i);
+
+ idpf_tx_buf_next(tx_buf) = next_buf_id;
+ } else {
+ next_buf_id = i;
+ }
+ tx_buf = &tx_q->tx_buf[next_buf_id];
+
size = skb_frag_size(frag);
data_len -= size;
@@ -2613,6 +2630,7 @@ static void idpf_tx_splitq_map(struct idpf_tx_queue *tx_q,
/* write last descriptor with RS and EOP bits */
first->rs_idx = i;
+ idpf_tx_buf_next(tx_buf) = IDPF_TXBUF_NULL;
td_cmd |= params->eop_cmd;
idpf_tx_splitq_build_desc(tx_desc, params, td_cmd, size);
i = idpf_tx_splitq_bump_ntu(tx_q, i);
@@ -2821,8 +2839,6 @@ idpf_tx_splitq_get_ctx_desc(struct idpf_tx_queue *txq)
union idpf_flex_tx_ctx_desc *desc;
int i = txq->next_to_use;
- txq->tx_buf[i].type = LIBETH_SQE_CTX;
-
/* grab the next descriptor */
desc = &txq->flex_ctx[i];
txq->next_to_use = idpf_tx_splitq_bump_ntu(txq, i);
@@ -2945,6 +2961,7 @@ static netdev_tx_t idpf_tx_splitq_frame(struct sk_buff *skb,
struct idpf_tx_buf *first;
unsigned int count;
int tso, idx;
+ u16 buf_id;
count = idpf_tx_desc_count_required(tx_q, skb);
if (unlikely(!count))
@@ -2988,23 +3005,11 @@ static netdev_tx_t idpf_tx_splitq_frame(struct sk_buff *skb,
idpf_tx_set_tstamp_desc(ctx_desc, idx);
}
- /* record the location of the first descriptor for this packet */
- first = &tx_q->tx_buf[tx_q->next_to_use];
- first->skb = skb;
-
- if (tso) {
- first->packets = tx_params.offload.tso_segs;
- first->bytes = skb->len +
- ((first->packets - 1) * tx_params.offload.tso_hdr_len);
- } else {
- first->packets = 1;
- first->bytes = max_t(unsigned int, skb->len, ETH_ZLEN);
- }
-
if (idpf_queue_has(FLOW_SCH_EN, tx_q)) {
if (unlikely(!idpf_tx_get_free_buf_id(tx_q->refillq,
&tx_params.compl_tag)))
return idpf_tx_drop_skb(tx_q, skb);
+ buf_id = tx_params.compl_tag;
tx_params.dtype = IDPF_TX_DESC_DTYPE_FLEX_FLOW_SCHE;
tx_params.eop_cmd = IDPF_TXD_FLEX_FLOW_CMD_EOP;
@@ -3023,6 +3028,8 @@ static netdev_tx_t idpf_tx_splitq_frame(struct sk_buff *skb,
tx_params.offload.td_cmd |= IDPF_TXD_FLEX_FLOW_CMD_CS_EN;
} else {
+ buf_id = tx_q->next_to_use;
+
tx_params.dtype = IDPF_TX_DESC_DTYPE_FLEX_L2TAG1_L2TAG2;
tx_params.eop_cmd = IDPF_TXD_LAST_DESC_CMD;
@@ -3030,6 +3037,18 @@ static netdev_tx_t idpf_tx_splitq_frame(struct sk_buff *skb,
tx_params.offload.td_cmd |= IDPF_TX_FLEX_DESC_CMD_CS_EN;
}
+ first = &tx_q->tx_buf[buf_id];
+ first->skb = skb;
+
+ if (tso) {
+ first->packets = tx_params.offload.tso_segs;
+ first->bytes = skb->len +
+ ((first->packets - 1) * tx_params.offload.tso_hdr_len);
+ } else {
+ first->packets = 1;
+ first->bytes = max_t(unsigned int, skb->len, ETH_ZLEN);
+ }
+
idpf_tx_splitq_map(tx_q, &tx_params, first);
return NETDEV_TX_OK;
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.h b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
index 65acad4c929d..a79a6a89c5e3 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.h
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
@@ -136,6 +136,8 @@ do { \
((++(txq)->compl_tag_cur_gen) >= (txq)->compl_tag_gen_max ? \
0 : (txq)->compl_tag_cur_gen)
+#define IDPF_TXBUF_NULL U32_MAX
+
#define IDPF_TXD_LAST_DESC_CMD (IDPF_TX_DESC_CMD_EOP | IDPF_TX_DESC_CMD_RS)
#define IDPF_TX_FLAGS_TSO BIT(0)
@@ -634,6 +636,7 @@ libeth_cacheline_set_assert(struct idpf_rx_queue, 64,
* @size: Length of descriptor ring in bytes
* @dma: Physical address of ring
* @q_vector: Backreference to associated vector
+ * @buf_pool_size: Total number of idpf_tx_buf
*/
struct idpf_tx_queue {
__cacheline_group_begin_aligned(read_mostly);
@@ -692,11 +695,12 @@ struct idpf_tx_queue {
dma_addr_t dma;
struct idpf_q_vector *q_vector;
+ u32 buf_pool_size;
__cacheline_group_end_aligned(cold);
};
libeth_cacheline_set_assert(struct idpf_tx_queue, 64,
120 + sizeof(struct u64_stats_sync),
- 24);
+ 32);
/**
* struct idpf_buf_queue - software structure representing a buffer queue
@@ -1041,8 +1045,8 @@ void idpf_tx_buf_hw_update(struct idpf_tx_queue *tx_q, u32 val,
bool xmit_more);
unsigned int idpf_size_to_txd_count(unsigned int size);
netdev_tx_t idpf_tx_drop_skb(struct idpf_tx_queue *tx_q, struct sk_buff *skb);
-void idpf_tx_dma_map_error(struct idpf_tx_queue *txq, struct sk_buff *skb,
- struct idpf_tx_buf *first, u16 ring_idx);
+void idpf_tx_dma_map_unwind(struct idpf_tx_queue *txq, struct sk_buff *skb,
+ struct idpf_tx_buf *first, u16 ring_idx);
unsigned int idpf_tx_desc_count_required(struct idpf_tx_queue *txq,
struct sk_buff *skb);
void idpf_tx_timeout(struct net_device *netdev, unsigned int txqueue);
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread* Re: [Intel-wired-lan] [PATCH net 3/5] idpf: replace flow scheduling buffer ring with buffer pool
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 3/5] idpf: replace flow scheduling buffer ring with buffer pool Joshua Hay
@ 2025-06-27 13:10 ` Alexander Lobakin
2025-06-30 16:11 ` Hay, Joshua A
0 siblings, 1 reply; 16+ messages in thread
From: Alexander Lobakin @ 2025-06-27 13:10 UTC (permalink / raw)
To: Joshua Hay
Cc: intel-wired-lan, netdev, Luigi Rizzo, Brian Vazquez,
Madhu Chittim
From: Joshua Hay <joshua.a.hay@intel.com>
Date: Wed, 25 Jun 2025 09:11:54 -0700
> Replace the TxQ buffer ring with one large pool/array of buffers (only
> for flow scheduling). The completion tag passed to HW through the
[...]
> diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.c b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
> index cdecf558d7ec..25eea632a966 100644
> --- a/drivers/net/ethernet/intel/idpf/idpf_txrx.c
> +++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
> @@ -13,6 +13,7 @@ struct idpf_tx_stash {
> struct libeth_sqe buf;
> };
>
> +#define idpf_tx_buf_next(buf) (*(u32 *)&(buf)->priv)
Align it to the next line, i.e. 2 tabs instead of 2 spaces.
> #define idpf_tx_buf_compl_tag(buf) (*(u32 *)&(buf)->priv)
> LIBETH_SQE_CHECK_PRIV(u32);
>
> @@ -91,7 +92,7 @@ static void idpf_tx_buf_rel_all(struct idpf_tx_queue *txq)
> return;
>
> /* Free all the Tx buffer sk_buffs */
> - for (i = 0; i < txq->desc_count; i++)
> + for (i = 0; i < txq->buf_pool_size; i++)
> libeth_tx_complete(&txq->tx_buf[i], &cp);
>
> kfree(txq->tx_buf);
> @@ -205,7 +206,11 @@ static int idpf_tx_buf_alloc_all(struct idpf_tx_queue *tx_q)
> /* Allocate book keeping buffers only. Buffers to be supplied to HW
> * are allocated by kernel network stack and received as part of skb
> */
> - buf_size = sizeof(struct idpf_tx_buf) * tx_q->desc_count;
> + if (idpf_queue_has(FLOW_SCH_EN, tx_q))
> + tx_q->buf_pool_size = U16_MAX;
3.2 Mb per queue... OTOH 1 Rx queue with 512 descriptors eats 2.1 Mb,
not that bad.
> + else
> + tx_q->buf_pool_size = tx_q->desc_count;
> + buf_size = sizeof(struct idpf_tx_buf) * tx_q->buf_pool_size;
array_size() if you really want, but the proper way would be to replace
the kzalloc() below with kcalloc().
> tx_q->tx_buf = kzalloc(buf_size, GFP_KERNEL);
> if (!tx_q->tx_buf)
> return -ENOMEM;
[...]
> +static bool idpf_tx_clean_bufs(struct idpf_tx_queue *txq, u16 buf_id,
Just use u32 when it comes to function arguments and onstack variables.
> + struct libeth_sq_napi_stats *cleaned,
> + int budget)
> {
> - u16 idx = compl_tag & txq->compl_tag_bufid_m;
> + u16 idx = buf_id & txq->compl_tag_bufid_m;
> struct idpf_tx_buf *tx_buf = NULL;
> struct libeth_cq_pp cp = {
> .dev = txq->dev,
[...]
> if (idpf_queue_has(FLOW_SCH_EN, tx_q)) {
> if (unlikely(!idpf_tx_get_free_buf_id(tx_q->refillq,
> &tx_params.compl_tag)))
> return idpf_tx_drop_skb(tx_q, skb);
> + buf_id = tx_params.compl_tag;
So this field in tx_params needs to be renamed as it no longer reflects
its purpose.
>
> tx_params.dtype = IDPF_TX_DESC_DTYPE_FLEX_FLOW_SCHE;
> tx_params.eop_cmd = IDPF_TXD_FLEX_FLOW_CMD_EOP;
Thanks,
Olek
^ permalink raw reply [flat|nested] 16+ messages in thread* RE: [Intel-wired-lan] [PATCH net 3/5] idpf: replace flow scheduling buffer ring with buffer pool
2025-06-27 13:10 ` Alexander Lobakin
@ 2025-06-30 16:11 ` Hay, Joshua A
0 siblings, 0 replies; 16+ messages in thread
From: Hay, Joshua A @ 2025-06-30 16:11 UTC (permalink / raw)
To: Lobakin, Aleksander
Cc: intel-wired-lan@lists.osuosl.org, netdev@vger.kernel.org,
Luigi Rizzo, Brian Vazquez, Chittim, Madhu
> > Replace the TxQ buffer ring with one large pool/array of buffers (only
> > for flow scheduling). The completion tag passed to HW through the
>
> [...]
>
> > diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.c
> b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
> > index cdecf558d7ec..25eea632a966 100644
> > --- a/drivers/net/ethernet/intel/idpf/idpf_txrx.c
> > +++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
> > @@ -13,6 +13,7 @@ struct idpf_tx_stash {
> > struct libeth_sqe buf;
> > };
> >
> > +#define idpf_tx_buf_next(buf) (*(u32 *)&(buf)->priv)
>
> Align it to the next line, i.e. 2 tabs instead of 2 spaces.
Will fix in v2
>
> > #define idpf_tx_buf_compl_tag(buf) (*(u32 *)&(buf)->priv)
> > LIBETH_SQE_CHECK_PRIV(u32);
> >
> > @@ -91,7 +92,7 @@ static void idpf_tx_buf_rel_all(struct idpf_tx_queue
> *txq)
> > return;
> >
> > /* Free all the Tx buffer sk_buffs */
> > - for (i = 0; i < txq->desc_count; i++)
> > + for (i = 0; i < txq->buf_pool_size; i++)
> > libeth_tx_complete(&txq->tx_buf[i], &cp);
> >
> > kfree(txq->tx_buf);
> > @@ -205,7 +206,11 @@ static int idpf_tx_buf_alloc_all(struct
> idpf_tx_queue *tx_q)
> > /* Allocate book keeping buffers only. Buffers to be supplied to HW
> > * are allocated by kernel network stack and received as part of skb
> > */
> > - buf_size = sizeof(struct idpf_tx_buf) * tx_q->desc_count;
> > + if (idpf_queue_has(FLOW_SCH_EN, tx_q))
> > + tx_q->buf_pool_size = U16_MAX;
>
> 3.2 Mb per queue... OTOH 1 Rx queue with 512 descriptors eats 2.1 Mb,
> not that bad.
>
> > + else
> > + tx_q->buf_pool_size = tx_q->desc_count;
> > + buf_size = sizeof(struct idpf_tx_buf) * tx_q->buf_pool_size;
>
> array_size() if you really want, but the proper way would be to replace
> the kzalloc() below with kcalloc().
Will fix in v2
>
> > tx_q->tx_buf = kzalloc(buf_size, GFP_KERNEL);
> > if (!tx_q->tx_buf)
> > return -ENOMEM;
>
> [...]
>
> > +static bool idpf_tx_clean_bufs(struct idpf_tx_queue *txq, u16 buf_id,
>
> Just use u32 when it comes to function arguments and onstack variables.
Will fix in v2
>
> > + struct libeth_sq_napi_stats *cleaned,
> > + int budget)
> > {
> > - u16 idx = compl_tag & txq->compl_tag_bufid_m;
> > + u16 idx = buf_id & txq->compl_tag_bufid_m;
> > struct idpf_tx_buf *tx_buf = NULL;
> > struct libeth_cq_pp cp = {
> > .dev = txq->dev,
>
> [...]
>
> > if (idpf_queue_has(FLOW_SCH_EN, tx_q)) {
> > if (unlikely(!idpf_tx_get_free_buf_id(tx_q->refillq,
> > &tx_params.compl_tag)))
> > return idpf_tx_drop_skb(tx_q, skb);
> > + buf_id = tx_params.compl_tag;
>
> So this field in tx_params needs to be renamed as it no longer reflects
> its purpose.
I would prefer to keep the compl_tag name. The params struct represents what goes into the descriptors and from a descriptor perspective, its purpose has not changed. The compl_tag must be the same across all descriptors for a given packet. But we can use multiple buf_ids per packet.
However, I can swap it so we pass &buf_id to idpf_tx_get_free_buf_id and assign it to tx_params.compl_tag.
>
> >
> > tx_params.dtype = IDPF_TX_DESC_DTYPE_FLEX_FLOW_SCHE;
> > tx_params.eop_cmd = IDPF_TXD_FLEX_FLOW_CMD_EOP;
>
> Thanks,
> Olek
Thanks,
Josh
^ permalink raw reply [flat|nested] 16+ messages in thread
* [Intel-wired-lan] [PATCH net 4/5] idpf: stop Tx if there are insufficient buffer resources
2025-06-25 16:11 [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool Joshua Hay
` (2 preceding siblings ...)
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 3/5] idpf: replace flow scheduling buffer ring with buffer pool Joshua Hay
@ 2025-06-25 16:11 ` Joshua Hay
2025-06-27 13:17 ` Alexander Lobakin
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 5/5] idpf: remove obsolete stashing code Joshua Hay
2025-06-25 22:32 ` [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool Paul Menzel
5 siblings, 1 reply; 16+ messages in thread
From: Joshua Hay @ 2025-06-25 16:11 UTC (permalink / raw)
To: intel-wired-lan; +Cc: netdev, Joshua Hay, Madhu Chittim
The Tx refillq logic will cause packets to be silently dropped if there
are not enough buffer resources available to send a packet in flow
scheduling mode. Instead, determine how many buffers are needed along
with number of descriptors. Make sure there are enough of both resources
to send the packet, and stop the queue if not.
Fixes: 7292af042bcf ("idpf: fix a race in txq wakeup")
Signed-off-by: Joshua Hay <joshua.a.hay@intel.com>
Reviewed-by: Madhu Chittim <madhu.chittim@intel.com>
---
.../ethernet/intel/idpf/idpf_singleq_txrx.c | 4 +-
drivers/net/ethernet/intel/idpf/idpf_txrx.c | 51 +++++++++++++------
drivers/net/ethernet/intel/idpf/idpf_txrx.h | 9 +++-
3 files changed, 44 insertions(+), 20 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c b/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c
index 4c1d7235a073..d8216bb13019 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c
@@ -361,12 +361,12 @@ netdev_tx_t idpf_tx_singleq_frame(struct sk_buff *skb,
struct idpf_tx_queue *tx_q)
{
struct idpf_tx_offload_params offload = { };
+ unsigned int count, buf_count;
struct idpf_tx_buf *first;
int csum, tso, needed;
- unsigned int count;
__be16 protocol;
- count = idpf_tx_desc_count_required(tx_q, skb);
+ count = idpf_tx_res_count_required(tx_q, skb, &buf_count);
if (unlikely(!count))
return idpf_tx_drop_skb(tx_q, skb);
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.c b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
index 25eea632a966..674b7a382bf1 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
@@ -2254,15 +2254,22 @@ void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc,
desc->flow.qw1.compl_tag = cpu_to_le16(params->compl_tag);
}
-/* Global conditions to tell whether the txq (and related resources)
- * has room to allow the use of "size" descriptors.
+/**
+ * idpf_tx_splitq_has_room - check if enough Tx splitq resources are available
+ * @tx_q: the queue to be checked
+ * @descs_needed: number of descriptors required for this packet
+ * @bufs_needed: number of Tx buffers required for this packet
+ *
+ * Return: 0 if no room available, 1 otherwise
*/
-static int idpf_txq_has_room(struct idpf_tx_queue *tx_q, u32 size)
+static int idpf_txq_has_room(struct idpf_tx_queue *tx_q, u32 descs_needed,
+ u32 bufs_needed)
{
- if (IDPF_DESC_UNUSED(tx_q) < size ||
+ if (IDPF_DESC_UNUSED(tx_q) < descs_needed ||
IDPF_TX_COMPLQ_PENDING(tx_q->txq_grp) >
IDPF_TX_COMPLQ_OVERFLOW_THRESH(tx_q->txq_grp->complq) ||
- IDPF_TX_BUF_RSV_LOW(tx_q))
+ IDPF_TX_BUF_RSV_LOW(tx_q) ||
+ IDPF_BUFS_UNUSED(tx_q->refillq) < bufs_needed)
return 0;
return 1;
}
@@ -2271,14 +2278,21 @@ static int idpf_txq_has_room(struct idpf_tx_queue *tx_q, u32 size)
* idpf_tx_maybe_stop_splitq - 1st level check for Tx splitq stop conditions
* @tx_q: the queue to be checked
* @descs_needed: number of descriptors required for this packet
+ * @bufs_needed: number of buffers needed for this packet
*
- * Returns 0 if stop is not needed
+ * Return: 0 if stop is not needed
*/
static int idpf_tx_maybe_stop_splitq(struct idpf_tx_queue *tx_q,
- unsigned int descs_needed)
+ unsigned int descs_needed,
+ unsigned int bufs_needed)
{
+ /* Since we have multiple resources to check for splitq, our
+ * start,stop_thrs becomes a boolean check instead of a count
+ * threshold.
+ */
if (netif_subqueue_maybe_stop(tx_q->netdev, tx_q->idx,
- idpf_txq_has_room(tx_q, descs_needed),
+ idpf_txq_has_room(tx_q, descs_needed,
+ bufs_needed),
1, 1))
return 0;
@@ -2320,24 +2334,29 @@ void idpf_tx_buf_hw_update(struct idpf_tx_queue *tx_q, u32 val,
}
/**
- * idpf_tx_desc_count_required - calculate number of Tx descriptors needed
+ * idpf_tx_res_count_required - get number of Tx resources needed for this pkt
* @txq: queue to send buffer on
* @skb: send buffer
+ * @bufs_needed: (output) number of buffers needed for this skb.
*
- * Returns number of data descriptors needed for this skb.
+ * Return: number of data descriptors and buffers needed for this skb.
*/
-unsigned int idpf_tx_desc_count_required(struct idpf_tx_queue *txq,
- struct sk_buff *skb)
+unsigned int idpf_tx_res_count_required(struct idpf_tx_queue *txq,
+ struct sk_buff *skb,
+ unsigned int *bufs_needed)
{
const struct skb_shared_info *shinfo;
unsigned int count = 0, i;
count += !!skb_headlen(skb);
- if (!skb_is_nonlinear(skb))
+ if (!skb_is_nonlinear(skb)) {
+ *bufs_needed = 1;
return count;
+ }
shinfo = skb_shinfo(skb);
+ *bufs_needed = shinfo->nr_frags;
for (i = 0; i < shinfo->nr_frags; i++) {
unsigned int size;
@@ -2958,12 +2977,12 @@ static netdev_tx_t idpf_tx_splitq_frame(struct sk_buff *skb,
{
struct idpf_tx_splitq_params tx_params = { };
union idpf_flex_tx_ctx_desc *ctx_desc;
+ unsigned int count, buf_count;
struct idpf_tx_buf *first;
- unsigned int count;
int tso, idx;
u16 buf_id;
- count = idpf_tx_desc_count_required(tx_q, skb);
+ count = idpf_tx_res_count_required(tx_q, skb, &buf_count);
if (unlikely(!count))
return idpf_tx_drop_skb(tx_q, skb);
@@ -2973,7 +2992,7 @@ static netdev_tx_t idpf_tx_splitq_frame(struct sk_buff *skb,
/* Check for splitq specific TX resources */
count += (IDPF_TX_DESCS_PER_CACHE_LINE + tso);
- if (idpf_tx_maybe_stop_splitq(tx_q, count)) {
+ if (idpf_tx_maybe_stop_splitq(tx_q, count, buf_count)) {
idpf_tx_buf_hw_update(tx_q, tx_q->next_to_use, false);
return NETDEV_TX_BUSY;
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.h b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
index a79a6a89c5e3..2d4846793f5a 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.h
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
@@ -116,6 +116,10 @@ do { \
#define IDPF_DESC_UNUSED(txq) \
((((txq)->next_to_clean > (txq)->next_to_use) ? 0 : (txq)->desc_count) + \
(txq)->next_to_clean - (txq)->next_to_use - 1)
+#define IDPF_BUFS_UNUSED(refillq) \
+ ((((refillq)->next_to_use > (refillq)->next_to_clean) ? \
+ 0 : (refillq)->desc_count) + \
+ (refillq)->next_to_use - (refillq)->next_to_clean)
#define IDPF_TX_BUF_RSV_UNUSED(txq) ((txq)->stash->buf_stack.top)
#define IDPF_TX_BUF_RSV_LOW(txq) (IDPF_TX_BUF_RSV_UNUSED(txq) < \
@@ -1047,8 +1051,9 @@ unsigned int idpf_size_to_txd_count(unsigned int size);
netdev_tx_t idpf_tx_drop_skb(struct idpf_tx_queue *tx_q, struct sk_buff *skb);
void idpf_tx_dma_map_unwind(struct idpf_tx_queue *txq, struct sk_buff *skb,
struct idpf_tx_buf *first, u16 ring_idx);
-unsigned int idpf_tx_desc_count_required(struct idpf_tx_queue *txq,
- struct sk_buff *skb);
+unsigned int idpf_tx_res_count_required(struct idpf_tx_queue *txq,
+ struct sk_buff *skb,
+ unsigned int *buf_count);
void idpf_tx_timeout(struct net_device *netdev, unsigned int txqueue);
netdev_tx_t idpf_tx_singleq_frame(struct sk_buff *skb,
struct idpf_tx_queue *tx_q);
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread* Re: [Intel-wired-lan] [PATCH net 4/5] idpf: stop Tx if there are insufficient buffer resources
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 4/5] idpf: stop Tx if there are insufficient buffer resources Joshua Hay
@ 2025-06-27 13:17 ` Alexander Lobakin
0 siblings, 0 replies; 16+ messages in thread
From: Alexander Lobakin @ 2025-06-27 13:17 UTC (permalink / raw)
To: Joshua Hay; +Cc: intel-wired-lan, netdev, Madhu Chittim
From: Joshua Hay <joshua.a.hay@intel.com>
Date: Wed, 25 Jun 2025 09:11:55 -0700
> The Tx refillq logic will cause packets to be silently dropped if there
> are not enough buffer resources available to send a packet in flow
> scheduling mode. Instead, determine how many buffers are needed along
> with number of descriptors. Make sure there are enough of both resources
> to send the packet, and stop the queue if not.
>
> Fixes: 7292af042bcf ("idpf: fix a race in txq wakeup")
> Signed-off-by: Joshua Hay <joshua.a.hay@intel.com>
> Reviewed-by: Madhu Chittim <madhu.chittim@intel.com>
> ---
> .../ethernet/intel/idpf/idpf_singleq_txrx.c | 4 +-
> drivers/net/ethernet/intel/idpf/idpf_txrx.c | 51 +++++++++++++------
> drivers/net/ethernet/intel/idpf/idpf_txrx.h | 9 +++-
> 3 files changed, 44 insertions(+), 20 deletions(-)
>
> diff --git a/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c b/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c
> index 4c1d7235a073..d8216bb13019 100644
> --- a/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c
> +++ b/drivers/net/ethernet/intel/idpf/idpf_singleq_txrx.c
> @@ -361,12 +361,12 @@ netdev_tx_t idpf_tx_singleq_frame(struct sk_buff *skb,
> struct idpf_tx_queue *tx_q)
> {
> struct idpf_tx_offload_params offload = { };
> + unsigned int count, buf_count;
u32
> struct idpf_tx_buf *first;
> int csum, tso, needed;
> - unsigned int count;
> __be16 protocol;
>
> - count = idpf_tx_desc_count_required(tx_q, skb);
> + count = idpf_tx_res_count_required(tx_q, skb, &buf_count);
> if (unlikely(!count))
> return idpf_tx_drop_skb(tx_q, skb);
[...]
> static int idpf_tx_maybe_stop_splitq(struct idpf_tx_queue *tx_q,
> - unsigned int descs_needed)
> + unsigned int descs_needed,
> + unsigned int bufs_needed)
Same (and everywhere else).
> {
> + /* Since we have multiple resources to check for splitq, our
> + * start,stop_thrs becomes a boolean check instead of a count
> + * threshold.
> + */
> if (netif_subqueue_maybe_stop(tx_q->netdev, tx_q->idx,
> - idpf_txq_has_room(tx_q, descs_needed),
> + idpf_txq_has_room(tx_q, descs_needed,
> + bufs_needed),
> 1, 1))
> return 0;
[...]
> diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.h b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
> index a79a6a89c5e3..2d4846793f5a 100644
> --- a/drivers/net/ethernet/intel/idpf/idpf_txrx.h
> +++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
> @@ -116,6 +116,10 @@ do { \
> #define IDPF_DESC_UNUSED(txq) \
> ((((txq)->next_to_clean > (txq)->next_to_use) ? 0 : (txq)->desc_count) + \
> (txq)->next_to_clean - (txq)->next_to_use - 1)
> +#define IDPF_BUFS_UNUSED(refillq) \
> + ((((refillq)->next_to_use > (refillq)->next_to_clean) ? \
> + 0 : (refillq)->desc_count) + \
> + (refillq)->next_to_use - (refillq)->next_to_clean)
Just make a static inline.
>
> #define IDPF_TX_BUF_RSV_UNUSED(txq) ((txq)->stash->buf_stack.top)
> #define IDPF_TX_BUF_RSV_LOW(txq) (IDPF_TX_BUF_RSV_UNUSED(txq) < \
Thanks,
Olek
^ permalink raw reply [flat|nested] 16+ messages in thread
* [Intel-wired-lan] [PATCH net 5/5] idpf: remove obsolete stashing code
2025-06-25 16:11 [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool Joshua Hay
` (3 preceding siblings ...)
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 4/5] idpf: stop Tx if there are insufficient buffer resources Joshua Hay
@ 2025-06-25 16:11 ` Joshua Hay
2025-06-25 22:32 ` [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool Paul Menzel
5 siblings, 0 replies; 16+ messages in thread
From: Joshua Hay @ 2025-06-25 16:11 UTC (permalink / raw)
To: intel-wired-lan; +Cc: netdev, Joshua Hay, Madhu Chittim
With the new Tx buffer management scheme, there is no need for all of
the stashing mechanisms, the hash table, the reserve buffer stack, etc.
Remove all of that.
Signed-off-by: Joshua Hay <joshua.a.hay@intel.com>
Reviewed-by: Madhu Chittim <madhu.chittim@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf_txrx.c | 375 ++------------------
drivers/net/ethernet/intel/idpf/idpf_txrx.h | 47 +--
2 files changed, 23 insertions(+), 399 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.c b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
index 674b7a382bf1..811fc6d124a7 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
@@ -8,48 +8,12 @@
#include "idpf_ptp.h"
#include "idpf_virtchnl.h"
-struct idpf_tx_stash {
- struct hlist_node hlist;
- struct libeth_sqe buf;
-};
-
#define idpf_tx_buf_next(buf) (*(u32 *)&(buf)->priv)
-#define idpf_tx_buf_compl_tag(buf) (*(u32 *)&(buf)->priv)
LIBETH_SQE_CHECK_PRIV(u32);
static bool idpf_chk_linearize(struct sk_buff *skb, unsigned int max_bufs,
unsigned int count);
-/**
- * idpf_buf_lifo_push - push a buffer pointer onto stack
- * @stack: pointer to stack struct
- * @buf: pointer to buf to push
- *
- * Returns 0 on success, negative on failure
- **/
-static int idpf_buf_lifo_push(struct idpf_buf_lifo *stack,
- struct idpf_tx_stash *buf)
-{
- if (unlikely(stack->top == stack->size))
- return -ENOSPC;
-
- stack->bufs[stack->top++] = buf;
-
- return 0;
-}
-
-/**
- * idpf_buf_lifo_pop - pop a buffer pointer from stack
- * @stack: pointer to stack struct
- **/
-static struct idpf_tx_stash *idpf_buf_lifo_pop(struct idpf_buf_lifo *stack)
-{
- if (unlikely(!stack->top))
- return NULL;
-
- return stack->bufs[--stack->top];
-}
-
/**
* idpf_tx_timeout - Respond to a Tx Hang
* @netdev: network interface device structure
@@ -78,14 +42,11 @@ void idpf_tx_timeout(struct net_device *netdev, unsigned int txqueue)
static void idpf_tx_buf_rel_all(struct idpf_tx_queue *txq)
{
struct libeth_sq_napi_stats ss = { };
- struct idpf_buf_lifo *buf_stack;
- struct idpf_tx_stash *stash;
struct libeth_cq_pp cp = {
.dev = txq->dev,
.ss = &ss,
};
- struct hlist_node *tmp;
- u32 i, tag;
+ u32 i;
/* Buffers already cleared, nothing to do */
if (!txq->tx_buf)
@@ -97,33 +58,6 @@ static void idpf_tx_buf_rel_all(struct idpf_tx_queue *txq)
kfree(txq->tx_buf);
txq->tx_buf = NULL;
-
- if (!idpf_queue_has(FLOW_SCH_EN, txq))
- return;
-
- buf_stack = &txq->stash->buf_stack;
- if (!buf_stack->bufs)
- return;
-
- /*
- * If a Tx timeout occurred, there are potentially still bufs in the
- * hash table, free them here.
- */
- hash_for_each_safe(txq->stash->sched_buf_hash, tag, tmp, stash,
- hlist) {
- if (!stash)
- continue;
-
- libeth_tx_complete(&stash->buf, &cp);
- hash_del(&stash->hlist);
- idpf_buf_lifo_push(buf_stack, stash);
- }
-
- for (i = 0; i < buf_stack->size; i++)
- kfree(buf_stack->bufs[i]);
-
- kfree(buf_stack->bufs);
- buf_stack->bufs = NULL;
}
/**
@@ -199,9 +133,7 @@ static void idpf_tx_desc_rel_all(struct idpf_vport *vport)
*/
static int idpf_tx_buf_alloc_all(struct idpf_tx_queue *tx_q)
{
- struct idpf_buf_lifo *buf_stack;
int buf_size;
- int i;
/* Allocate book keeping buffers only. Buffers to be supplied to HW
* are allocated by kernel network stack and received as part of skb
@@ -215,29 +147,6 @@ static int idpf_tx_buf_alloc_all(struct idpf_tx_queue *tx_q)
if (!tx_q->tx_buf)
return -ENOMEM;
- if (!idpf_queue_has(FLOW_SCH_EN, tx_q))
- return 0;
-
- buf_stack = &tx_q->stash->buf_stack;
-
- /* Initialize tx buf stack for out-of-order completions if
- * flow scheduling offload is enabled
- */
- buf_stack->bufs = kcalloc(tx_q->desc_count, sizeof(*buf_stack->bufs),
- GFP_KERNEL);
- if (!buf_stack->bufs)
- return -ENOMEM;
-
- buf_stack->size = tx_q->desc_count;
- buf_stack->top = tx_q->desc_count;
-
- for (i = 0; i < tx_q->desc_count; i++) {
- buf_stack->bufs[i] = kzalloc(sizeof(*buf_stack->bufs[i]),
- GFP_KERNEL);
- if (!buf_stack->bufs[i])
- return -ENOMEM;
- }
-
return 0;
}
@@ -352,8 +261,6 @@ static int idpf_tx_desc_alloc_all(struct idpf_vport *vport)
for (i = 0; i < vport->num_txq_grp; i++) {
for (j = 0; j < vport->txq_grps[i].num_txq; j++) {
struct idpf_tx_queue *txq = vport->txq_grps[i].txqs[j];
- u8 gen_bits = 0;
- u16 bufidx_mask;
err = idpf_tx_desc_alloc(vport, txq);
if (err) {
@@ -362,34 +269,6 @@ static int idpf_tx_desc_alloc_all(struct idpf_vport *vport)
i);
goto err_out;
}
-
- if (!idpf_is_queue_model_split(vport->txq_model))
- continue;
-
- txq->compl_tag_cur_gen = 0;
-
- /* Determine the number of bits in the bufid
- * mask and add one to get the start of the
- * generation bits
- */
- bufidx_mask = txq->desc_count - 1;
- while (bufidx_mask >> 1) {
- txq->compl_tag_gen_s++;
- bufidx_mask = bufidx_mask >> 1;
- }
- txq->compl_tag_gen_s++;
-
- gen_bits = IDPF_TX_SPLITQ_COMPL_TAG_WIDTH -
- txq->compl_tag_gen_s;
- txq->compl_tag_gen_max = GETMAXVAL(gen_bits);
-
- /* Set bufid mask based on location of first
- * gen bit; it cannot simply be the descriptor
- * ring size-1 since we can have size values
- * where not all of those bits are set.
- */
- txq->compl_tag_bufid_m =
- GETMAXVAL(txq->compl_tag_gen_s);
}
if (!idpf_is_queue_model_split(vport->txq_model))
@@ -1044,9 +923,6 @@ static void idpf_txq_group_rel(struct idpf_vport *vport)
kfree(txq_grp->complq);
txq_grp->complq = NULL;
-
- if (flow_sch_en)
- kfree(txq_grp->stashes);
}
kfree(vport->txq_grps);
vport->txq_grps = NULL;
@@ -1407,7 +1283,6 @@ static int idpf_txq_group_alloc(struct idpf_vport *vport, u16 num_txq)
for (i = 0; i < vport->num_txq_grp; i++) {
struct idpf_txq_group *tx_qgrp = &vport->txq_grps[i];
struct idpf_adapter *adapter = vport->adapter;
- struct idpf_txq_stash *stashes;
int j;
tx_qgrp->vport = vport;
@@ -1420,15 +1295,6 @@ static int idpf_txq_group_alloc(struct idpf_vport *vport, u16 num_txq)
goto err_alloc;
}
- if (split && flow_sch_en) {
- stashes = kcalloc(num_txq, sizeof(*stashes),
- GFP_KERNEL);
- if (!stashes)
- goto err_alloc;
-
- tx_qgrp->stashes = stashes;
- }
-
for (j = 0; j < tx_qgrp->num_txq; j++) {
struct idpf_tx_queue *q = tx_qgrp->txqs[j];
@@ -1448,11 +1314,6 @@ static int idpf_txq_group_alloc(struct idpf_vport *vport, u16 num_txq)
if (!flow_sch_en)
continue;
- if (split) {
- q->stash = &stashes[j];
- hash_init(q->stash->sched_buf_hash);
- }
-
idpf_queue_set(FLOW_SCH_EN, q);
q->refillq = kzalloc(sizeof(*q->refillq), GFP_KERNEL);
@@ -1744,87 +1605,6 @@ static void idpf_tx_read_tstamp(struct idpf_tx_queue *txq, struct sk_buff *skb)
spin_unlock_bh(&tx_tstamp_caps->status_lock);
}
-/**
- * idpf_tx_clean_stashed_bufs - clean bufs that were stored for
- * out of order completions
- * @txq: queue to clean
- * @compl_tag: completion tag of packet to clean (from completion descriptor)
- * @cleaned: pointer to stats struct to track cleaned packets/bytes
- * @budget: Used to determine if we are in netpoll
- */
-static void idpf_tx_clean_stashed_bufs(struct idpf_tx_queue *txq,
- u16 compl_tag,
- struct libeth_sq_napi_stats *cleaned,
- int budget)
-{
- struct idpf_tx_stash *stash;
- struct hlist_node *tmp_buf;
- struct libeth_cq_pp cp = {
- .dev = txq->dev,
- .ss = cleaned,
- .napi = budget,
- };
-
- /* Buffer completion */
- hash_for_each_possible_safe(txq->stash->sched_buf_hash, stash, tmp_buf,
- hlist, compl_tag) {
- if (unlikely(idpf_tx_buf_compl_tag(&stash->buf) != compl_tag))
- continue;
-
- hash_del(&stash->hlist);
-
- if (stash->buf.type == LIBETH_SQE_SKB &&
- (skb_shinfo(stash->buf.skb)->tx_flags & SKBTX_IN_PROGRESS))
- idpf_tx_read_tstamp(txq, stash->buf.skb);
-
- libeth_tx_complete(&stash->buf, &cp);
-
- /* Push shadow buf back onto stack */
- idpf_buf_lifo_push(&txq->stash->buf_stack, stash);
- }
-}
-
-/**
- * idpf_stash_flow_sch_buffers - store buffer parameters info to be freed at a
- * later time (only relevant for flow scheduling mode)
- * @txq: Tx queue to clean
- * @tx_buf: buffer to store
- */
-static int idpf_stash_flow_sch_buffers(struct idpf_tx_queue *txq,
- struct idpf_tx_buf *tx_buf)
-{
- struct idpf_tx_stash *stash;
-
- if (unlikely(tx_buf->type <= LIBETH_SQE_CTX))
- return 0;
-
- stash = idpf_buf_lifo_pop(&txq->stash->buf_stack);
- if (unlikely(!stash)) {
- net_err_ratelimited("%s: No out-of-order TX buffers left!\n",
- netdev_name(txq->netdev));
-
- return -ENOMEM;
- }
-
- /* Store buffer params in shadow buffer */
- stash->buf.skb = tx_buf->skb;
- stash->buf.bytes = tx_buf->bytes;
- stash->buf.packets = tx_buf->packets;
- stash->buf.type = tx_buf->type;
- stash->buf.nr_frags = tx_buf->nr_frags;
- dma_unmap_addr_set(&stash->buf, dma, dma_unmap_addr(tx_buf, dma));
- dma_unmap_len_set(&stash->buf, len, dma_unmap_len(tx_buf, len));
- idpf_tx_buf_compl_tag(&stash->buf) = idpf_tx_buf_compl_tag(tx_buf);
-
- /* Add buffer to buf_hash table to be freed later */
- hash_add(txq->stash->sched_buf_hash, &stash->hlist,
- idpf_tx_buf_compl_tag(&stash->buf));
-
- tx_buf->type = LIBETH_SQE_EMPTY;
-
- return 0;
-}
-
#define idpf_tx_splitq_clean_bump_ntc(txq, ntc, desc, buf) \
do { \
if (unlikely(++(ntc) == (txq)->desc_count)) { \
@@ -1852,14 +1632,8 @@ do { \
* Separate packet completion events will be reported on the completion queue,
* and the buffers will be cleaned separately. The stats are not updated from
* this function when using flow-based scheduling.
- *
- * Furthermore, in flow scheduling mode, check to make sure there are enough
- * reserve buffers to stash the packet. If there are not, return early, which
- * will leave next_to_clean pointing to the packet that failed to be stashed.
- *
- * Return: false in the scenario above, true otherwise.
*/
-static bool idpf_tx_splitq_clean(struct idpf_tx_queue *tx_q, u16 end,
+static void idpf_tx_splitq_clean(struct idpf_tx_queue *tx_q, u16 end,
int napi_budget,
struct libeth_sq_napi_stats *cleaned,
bool descs_only)
@@ -1873,12 +1647,11 @@ static bool idpf_tx_splitq_clean(struct idpf_tx_queue *tx_q, u16 end,
.napi = napi_budget,
};
struct idpf_tx_buf *tx_buf;
- bool clean_complete = true;
if (descs_only) {
/* Bump ring index to mark as cleaned. */
tx_q->next_to_clean = end;
- return true;
+ return;
}
tx_desc = &tx_q->flex_tx[ntc];
@@ -1899,53 +1672,24 @@ static bool idpf_tx_splitq_clean(struct idpf_tx_queue *tx_q, u16 end,
break;
eop_idx = tx_buf->rs_idx;
+ libeth_tx_complete(tx_buf, &cp);
- if (descs_only) {
- if (IDPF_TX_BUF_RSV_UNUSED(tx_q) < tx_buf->nr_frags) {
- clean_complete = false;
- goto tx_splitq_clean_out;
- }
-
- idpf_stash_flow_sch_buffers(tx_q, tx_buf);
+ /* unmap remaining buffers */
+ while (ntc != eop_idx) {
+ idpf_tx_splitq_clean_bump_ntc(tx_q, ntc,
+ tx_desc, tx_buf);
- while (ntc != eop_idx) {
- idpf_tx_splitq_clean_bump_ntc(tx_q, ntc,
- tx_desc, tx_buf);
- idpf_stash_flow_sch_buffers(tx_q, tx_buf);
- }
- } else {
+ /* unmap any remaining paged data */
libeth_tx_complete(tx_buf, &cp);
-
- /* unmap remaining buffers */
- while (ntc != eop_idx) {
- idpf_tx_splitq_clean_bump_ntc(tx_q, ntc,
- tx_desc, tx_buf);
-
- /* unmap any remaining paged data */
- libeth_tx_complete(tx_buf, &cp);
- }
}
fetch_next_txq_desc:
idpf_tx_splitq_clean_bump_ntc(tx_q, ntc, tx_desc, tx_buf);
}
-tx_splitq_clean_out:
tx_q->next_to_clean = ntc;
-
- return clean_complete;
}
-#define idpf_tx_clean_buf_ring_bump_ntc(txq, ntc, buf) \
-do { \
- (buf)++; \
- (ntc)++; \
- if (unlikely((ntc) == (txq)->desc_count)) { \
- buf = (txq)->tx_buf; \
- ntc = 0; \
- } \
-} while (0)
-
/**
* idpf_tx_clean_bufs - clean flow scheduling TX queue buffers
* @txq: queue to clean
@@ -1956,18 +1700,16 @@ do { \
* Clean all buffers associated with the packet starting at buf_id. Returns the
* byte/segment count for the cleaned packet.
*/
-static bool idpf_tx_clean_bufs(struct idpf_tx_queue *txq, u16 buf_id,
+static void idpf_tx_clean_bufs(struct idpf_tx_queue *txq, u16 buf_id,
struct libeth_sq_napi_stats *cleaned,
int budget)
{
- u16 idx = buf_id & txq->compl_tag_bufid_m;
struct idpf_tx_buf *tx_buf = NULL;
struct libeth_cq_pp cp = {
.dev = txq->dev,
.ss = cleaned,
.napi = budget,
};
- u16 ntc, orig_idx = idx;
tx_buf = &txq->tx_buf[buf_id];
if (tx_buf->type == LIBETH_SQE_SKB) {
@@ -1985,67 +1727,6 @@ static bool idpf_tx_clean_bufs(struct idpf_tx_queue *txq, u16 buf_id,
libeth_tx_complete(tx_buf, &cp);
idpf_post_buf_refill(txq->refillq, buf_id);
}
-
- return true;
-
- tx_buf = &txq->tx_buf[idx];
-
- if (unlikely(tx_buf->type <= LIBETH_SQE_CTX ||
- idpf_tx_buf_compl_tag(tx_buf) != buf_id))
- return false;
-
- if (tx_buf->type == LIBETH_SQE_SKB) {
- if (skb_shinfo(tx_buf->skb)->tx_flags & SKBTX_IN_PROGRESS)
- idpf_tx_read_tstamp(txq, tx_buf->skb);
-
- libeth_tx_complete(tx_buf, &cp);
- }
-
- idpf_tx_clean_buf_ring_bump_ntc(txq, idx, tx_buf);
-
- while (idpf_tx_buf_compl_tag(tx_buf) == buf_id) {
- libeth_tx_complete(tx_buf, &cp);
- idpf_tx_clean_buf_ring_bump_ntc(txq, idx, tx_buf);
- }
-
- /*
- * It's possible the packet we just cleaned was an out of order
- * completion, which means we can stash the buffers starting from
- * the original next_to_clean and reuse the descriptors. We need
- * to compare the descriptor ring next_to_clean packet's "first" buffer
- * to the "first" buffer of the packet we just cleaned to determine if
- * this is the case. Howevever, next_to_clean can point to either a
- * reserved buffer that corresponds to a context descriptor used for the
- * next_to_clean packet (TSO packet) or the "first" buffer (single
- * packet). The orig_idx from the packet we just cleaned will always
- * point to the "first" buffer. If next_to_clean points to a reserved
- * buffer, let's bump ntc once and start the comparison from there.
- */
- ntc = txq->next_to_clean;
- tx_buf = &txq->tx_buf[ntc];
-
- if (tx_buf->type == LIBETH_SQE_CTX)
- idpf_tx_clean_buf_ring_bump_ntc(txq, ntc, tx_buf);
-
- /*
- * If ntc still points to a different "first" buffer, clean the
- * descriptor ring and stash all of the buffers for later cleaning. If
- * we cannot stash all of the buffers, next_to_clean will point to the
- * "first" buffer of the packet that could not be stashed and cleaning
- * will start there next time.
- */
- if (unlikely(tx_buf != &txq->tx_buf[orig_idx] &&
- !idpf_tx_splitq_clean(txq, orig_idx, budget, cleaned,
- true)))
- return true;
-
- /*
- * Otherwise, update next_to_clean to reflect the cleaning that was
- * done above.
- */
- txq->next_to_clean = idx;
-
- return true;
}
/**
@@ -2064,22 +1745,18 @@ static void idpf_tx_handle_rs_completion(struct idpf_tx_queue *txq,
struct libeth_sq_napi_stats *cleaned,
int budget)
{
- u16 compl_tag;
+ /*
+ * RS completion contains queue head for queue based scheduling or
+ * completion tag for flow based scheduling.
+ */
+ u16 rs_compl_val = le16_to_cpu(desc->q_head_compl_tag.q_head);
if (!idpf_queue_has(FLOW_SCH_EN, txq)) {
- u16 head = le16_to_cpu(desc->q_head_compl_tag.q_head);
-
- idpf_tx_splitq_clean(txq, head, budget, cleaned, false);
+ idpf_tx_splitq_clean(txq, rs_compl_val, budget, cleaned, false);
return;
}
- compl_tag = le16_to_cpu(desc->q_head_compl_tag.compl_tag);
-
- /* If we didn't clean anything on the ring, this packet must be
- * in the hash table. Go clean it there.
- */
- if (!idpf_tx_clean_bufs(txq, compl_tag, cleaned, budget))
- idpf_tx_clean_stashed_bufs(txq, compl_tag, cleaned, budget);
+ idpf_tx_clean_bufs(txq, rs_compl_val, cleaned, budget);
}
/**
@@ -2196,8 +1873,7 @@ static bool idpf_tx_clean_complq(struct idpf_compl_queue *complq, int budget,
/* Update BQL */
nq = netdev_get_tx_queue(tx_q->netdev, tx_q->idx);
- dont_wake = !complq_ok || IDPF_TX_BUF_RSV_LOW(tx_q) ||
- np->state != __IDPF_VPORT_UP ||
+ dont_wake = !complq_ok || np->state != __IDPF_VPORT_UP ||
!netif_carrier_ok(tx_q->netdev);
/* Check if the TXQ needs to and can be restarted */
__netif_txq_completed_wake(nq, tx_q->cleaned_pkts, tx_q->cleaned_bytes,
@@ -2268,7 +1944,6 @@ static int idpf_txq_has_room(struct idpf_tx_queue *tx_q, u32 descs_needed,
if (IDPF_DESC_UNUSED(tx_q) < descs_needed ||
IDPF_TX_COMPLQ_PENDING(tx_q->txq_grp) >
IDPF_TX_COMPLQ_OVERFLOW_THRESH(tx_q->txq_grp->complq) ||
- IDPF_TX_BUF_RSV_LOW(tx_q) ||
IDPF_BUFS_UNUSED(tx_q->refillq) < bufs_needed)
return 0;
return 1;
@@ -2445,10 +2120,8 @@ static unsigned int idpf_tx_splitq_bump_ntu(struct idpf_tx_queue *txq, u16 ntu)
{
ntu++;
- if (ntu == txq->desc_count) {
+ if (ntu == txq->desc_count)
ntu = 0;
- txq->compl_tag_cur_gen = IDPF_TX_ADJ_COMPL_TAG_GEN(txq);
- }
return ntu;
}
@@ -2587,8 +2260,6 @@ static void idpf_tx_splitq_map(struct idpf_tx_queue *tx_q,
if (unlikely(++i == tx_q->desc_count)) {
tx_desc = &tx_q->flex_tx[0];
i = 0;
- tx_q->compl_tag_cur_gen =
- IDPF_TX_ADJ_COMPL_TAG_GEN(tx_q);
} else {
tx_desc++;
}
@@ -2619,7 +2290,6 @@ static void idpf_tx_splitq_map(struct idpf_tx_queue *tx_q,
if (unlikely(++i == tx_q->desc_count)) {
tx_desc = &tx_q->flex_tx[0];
i = 0;
- tx_q->compl_tag_cur_gen = IDPF_TX_ADJ_COMPL_TAG_GEN(tx_q);
} else {
tx_desc++;
}
@@ -3032,10 +2702,9 @@ static netdev_tx_t idpf_tx_splitq_frame(struct sk_buff *skb,
tx_params.dtype = IDPF_TX_DESC_DTYPE_FLEX_FLOW_SCHE;
tx_params.eop_cmd = IDPF_TXD_FLEX_FLOW_CMD_EOP;
- /* Set the RE bit to catch any packets that may have not been
- * stashed during RS completion cleaning. MIN_GAP is set to
- * MIN_RING size to ensure it will be set at least once each
- * time around the ring.
+ /* Set the RE bit to periodically "clean" the descriptor ring.
+ * MIN_GAP is set to MIN_RING size to ensure it will be set at
+ * least once each time around the ring.
*/
if (idpf_tx_splitq_need_re(tx_q)) {
tx_params.eop_cmd |= IDPF_TXD_FLEX_FLOW_CMD_RE;
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.h b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
index 2d4846793f5a..7ec35a6c5a56 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.h
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
@@ -121,10 +121,6 @@ do { \
0 : (refillq)->desc_count) + \
(refillq)->next_to_use - (refillq)->next_to_clean)
-#define IDPF_TX_BUF_RSV_UNUSED(txq) ((txq)->stash->buf_stack.top)
-#define IDPF_TX_BUF_RSV_LOW(txq) (IDPF_TX_BUF_RSV_UNUSED(txq) < \
- (txq)->desc_count >> 2)
-
#define IDPF_TX_COMPLQ_OVERFLOW_THRESH(txcq) ((txcq)->desc_count >> 1)
/* Determine the absolute number of completions pending, i.e. the number of
* completions that are expected to arrive on the TX completion queue.
@@ -134,12 +130,6 @@ do { \
0 : U32_MAX) + \
(txq)->num_completions_pending - (txq)->complq->num_completions)
-#define IDPF_TX_SPLITQ_COMPL_TAG_WIDTH 16
-/* Adjust the generation for the completion tag and wrap if necessary */
-#define IDPF_TX_ADJ_COMPL_TAG_GEN(txq) \
- ((++(txq)->compl_tag_cur_gen) >= (txq)->compl_tag_gen_max ? \
- 0 : (txq)->compl_tag_cur_gen)
-
#define IDPF_TXBUF_NULL U32_MAX
#define IDPF_TXD_LAST_DESC_CMD (IDPF_TX_DESC_CMD_EOP | IDPF_TX_DESC_CMD_RS)
@@ -157,18 +147,6 @@ union idpf_tx_flex_desc {
#define idpf_tx_buf libeth_sqe
-/**
- * struct idpf_buf_lifo - LIFO for managing OOO completions
- * @top: Used to know how many buffers are left
- * @size: Total size of LIFO
- * @bufs: Backing array
- */
-struct idpf_buf_lifo {
- u16 top;
- u16 size;
- struct idpf_tx_stash **bufs;
-};
-
/**
* struct idpf_tx_offload_params - Offload parameters for a given packet
* @tx_flags: Feature flags enabled for this packet
@@ -472,17 +450,6 @@ struct idpf_tx_queue_stats {
#define IDPF_ITR_IDX_SPACING(spacing, dflt) (spacing ? spacing : dflt)
#define IDPF_DIM_DEFAULT_PROFILE_IX 1
-/**
- * struct idpf_txq_stash - Tx buffer stash for Flow-based scheduling mode
- * @buf_stack: Stack of empty buffers to store buffer info for out of order
- * buffer completions. See struct idpf_buf_lifo
- * @sched_buf_hash: Hash table to store buffers
- */
-struct idpf_txq_stash {
- struct idpf_buf_lifo buf_stack;
- DECLARE_HASHTABLE(sched_buf_hash, 12);
-} ____cacheline_aligned;
-
/**
* struct idpf_rx_queue - software structure representing a receive queue
* @rx: universal receive descriptor array
@@ -627,11 +594,7 @@ libeth_cacheline_set_assert(struct idpf_rx_queue, 64,
* only once at the end of the cleaning routine.
* @clean_budget: singleq only, queue cleaning budget
* @cleaned_pkts: Number of packets cleaned for the above said case
- * @stash: Tx buffer stash for Flow-based scheduling mode
* @refillq: Pointer to refill queue
- * @compl_tag_bufid_m: Completion tag buffer id mask
- * @compl_tag_cur_gen: Used to keep track of current completion tag generation
- * @compl_tag_gen_max: To determine when compl_tag_cur_gen should be reset
* @cached_tstamp_caps: Tx timestamp capabilities negotiated with the CP
* @tstamp_task: Work that handles Tx timestamp read
* @stats_sync: See struct u64_stats_sync
@@ -662,7 +625,6 @@ struct idpf_tx_queue {
u16 desc_count;
u16 tx_min_pkt_len;
- u16 compl_tag_gen_s;
struct net_device *netdev;
__cacheline_group_end_aligned(read_mostly);
@@ -679,13 +641,8 @@ struct idpf_tx_queue {
};
u16 cleaned_pkts;
- struct idpf_txq_stash *stash;
struct idpf_sw_queue *refillq;
- u16 compl_tag_bufid_m;
- u16 compl_tag_cur_gen;
- u16 compl_tag_gen_max;
-
struct idpf_ptp_vport_tx_tstamp_caps *cached_tstamp_caps;
struct work_struct *tstamp_task;
@@ -703,7 +660,7 @@ struct idpf_tx_queue {
__cacheline_group_end_aligned(cold);
};
libeth_cacheline_set_assert(struct idpf_tx_queue, 64,
- 120 + sizeof(struct u64_stats_sync),
+ 104 + sizeof(struct u64_stats_sync),
32);
/**
@@ -914,7 +871,6 @@ struct idpf_rxq_group {
* @vport: Vport back pointer
* @num_txq: Number of TX queues associated
* @txqs: Array of TX queue pointers
- * @stashes: array of OOO stashes for the queues
* @complq: Associated completion queue pointer, split queue only
* @num_completions_pending: Total number of completions pending for the
* completion queue, acculumated for all TX queues
@@ -929,7 +885,6 @@ struct idpf_txq_group {
u16 num_txq;
struct idpf_tx_queue *txqs[IDPF_LARGE_MAX_Q];
- struct idpf_txq_stash *stashes;
struct idpf_compl_queue *complq;
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread* Re: [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool
2025-06-25 16:11 [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool Joshua Hay
` (4 preceding siblings ...)
2025-06-25 16:11 ` [Intel-wired-lan] [PATCH net 5/5] idpf: remove obsolete stashing code Joshua Hay
@ 2025-06-25 22:32 ` Paul Menzel
2025-06-30 16:08 ` Hay, Joshua A
5 siblings, 1 reply; 16+ messages in thread
From: Paul Menzel @ 2025-06-25 22:32 UTC (permalink / raw)
To: Joshua Hay; +Cc: intel-wired-lan, netdev
Dear Joshua,
Thank you for these patches. One minor comment, should you resend.
Am 25.06.25 um 18:11 schrieb Joshua Hay:
> This series fixes a stability issue in the flow scheduling Tx send/clean
> path that results in a Tx timeout.
>
> The existing guardrails in the Tx path were not sufficient to prevent
> the driver from reusing completion tags that were still in flight (held
> by the HW). This collision would cause the driver to erroneously clean
> the wrong packet thus leaving the descriptor ring in a bad state.
>
> The main point of this refactor is replace the flow scheduling buffer
… to replace …?
> ring with a large pool/array of buffers. The completion tag then simply
> is the index into this array. The driver tracks the free tags and pulls
> the next free one from a refillq. The cleaning routines simply use the
> completion tag from the completion descriptor to index into the array to
> quickly find the buffers to clean.
>
> All of the code to support the refactor is added first to ensure traffic
> still passes with each patch. The final patch then removes all of the
> obsolete stashing code.
Do you have reproducers for the issue?
> Joshua Hay (5):
> idpf: add support for Tx refillqs in flow scheduling mode
> idpf: improve when to set RE bit logic
> idpf: replace flow scheduling buffer ring with buffer pool
> idpf: stop Tx if there are insufficient buffer resources
> idpf: remove obsolete stashing code
>
> .../ethernet/intel/idpf/idpf_singleq_txrx.c | 6 +-
> drivers/net/ethernet/intel/idpf/idpf_txrx.c | 626 ++++++------------
> drivers/net/ethernet/intel/idpf/idpf_txrx.h | 76 +--
> 3 files changed, 239 insertions(+), 469 deletions(-)
Kind regards,
Paul Menzel
^ permalink raw reply [flat|nested] 16+ messages in thread* RE: [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool
2025-06-25 22:32 ` [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool Paul Menzel
@ 2025-06-30 16:08 ` Hay, Joshua A
2025-06-30 16:22 ` Paul Menzel
0 siblings, 1 reply; 16+ messages in thread
From: Hay, Joshua A @ 2025-06-30 16:08 UTC (permalink / raw)
To: Paul Menzel; +Cc: intel-wired-lan@lists.osuosl.org, netdev@vger.kernel.org
> Dear Joshua,
>
>
> Thank you for these patches. One minor comment, should you resend.
>
> Am 25.06.25 um 18:11 schrieb Joshua Hay:
> > This series fixes a stability issue in the flow scheduling Tx send/clean
> > path that results in a Tx timeout.
> >
> > The existing guardrails in the Tx path were not sufficient to prevent
> > the driver from reusing completion tags that were still in flight (held
> > by the HW). This collision would cause the driver to erroneously clean
> > the wrong packet thus leaving the descriptor ring in a bad state.
> >
> > The main point of this refactor is replace the flow scheduling buffer
>
> … to replace …?
Thanks, will fix in v2
>
> > ring with a large pool/array of buffers. The completion tag then simply
> > is the index into this array. The driver tracks the free tags and pulls
> > the next free one from a refillq. The cleaning routines simply use the
> > completion tag from the completion descriptor to index into the array to
> > quickly find the buffers to clean.
> >
> > All of the code to support the refactor is added first to ensure traffic
> > still passes with each patch. The final patch then removes all of the
> > obsolete stashing code.
>
> Do you have reproducers for the issue?
This issue cannot be reproduced without the customer specific device configuration, but it can impact any traffic once in place.
>
> > Joshua Hay (5):
> > idpf: add support for Tx refillqs in flow scheduling mode
> > idpf: improve when to set RE bit logic
> > idpf: replace flow scheduling buffer ring with buffer pool
> > idpf: stop Tx if there are insufficient buffer resources
> > idpf: remove obsolete stashing code
> >
> > .../ethernet/intel/idpf/idpf_singleq_txrx.c | 6 +-
> > drivers/net/ethernet/intel/idpf/idpf_txrx.c | 626 ++++++------------
> > drivers/net/ethernet/intel/idpf/idpf_txrx.h | 76 +--
> > 3 files changed, 239 insertions(+), 469 deletions(-)
>
>
> Kind regards,
>
> Paul Menzel
Thanks,
Josh
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool
2025-06-30 16:08 ` Hay, Joshua A
@ 2025-06-30 16:22 ` Paul Menzel
2025-07-03 20:21 ` Hay, Joshua A
2025-07-07 14:43 ` Brian Vazquez
0 siblings, 2 replies; 16+ messages in thread
From: Paul Menzel @ 2025-06-30 16:22 UTC (permalink / raw)
To: Joshua A Hay; +Cc: intel-wired-lan, netdev
Dear Josh,
Am 30.06.25 um 18:08 schrieb Hay, Joshua A:
>> Am 25.06.25 um 18:11 schrieb Joshua Hay:
>>> This series fixes a stability issue in the flow scheduling Tx send/clean
>>> path that results in a Tx timeout.
>>>
>>> The existing guardrails in the Tx path were not sufficient to prevent
>>> the driver from reusing completion tags that were still in flight (held
>>> by the HW). This collision would cause the driver to erroneously clean
>>> the wrong packet thus leaving the descriptor ring in a bad state.
>>>
>>> The main point of this refactor is replace the flow scheduling buffer
>>
>> … to replace …?
>
> Thanks, will fix in v2
>
>>> ring with a large pool/array of buffers. The completion tag then simply
>>> is the index into this array. The driver tracks the free tags and pulls
>>> the next free one from a refillq. The cleaning routines simply use the
>>> completion tag from the completion descriptor to index into the array to
>>> quickly find the buffers to clean.
>>>
>>> All of the code to support the refactor is added first to ensure traffic
>>> still passes with each patch. The final patch then removes all of the
>>> obsolete stashing code.
>>
>> Do you have reproducers for the issue?
>
> This issue cannot be reproduced without the customer specific device
> configuration, but it can impact any traffic once in place.
Interesting. Then it’d be great if you could describe that setup in more
detail.
>>> Joshua Hay (5):
>>> idpf: add support for Tx refillqs in flow scheduling mode
>>> idpf: improve when to set RE bit logic
>>> idpf: replace flow scheduling buffer ring with buffer pool
>>> idpf: stop Tx if there are insufficient buffer resources
>>> idpf: remove obsolete stashing code
>>>
>>> .../ethernet/intel/idpf/idpf_singleq_txrx.c | 6 +-
>>> drivers/net/ethernet/intel/idpf/idpf_txrx.c | 626 ++++++------------
>>> drivers/net/ethernet/intel/idpf/idpf_txrx.h | 76 +--
>>> 3 files changed, 239 insertions(+), 469 deletions(-)
Kind regards,
Paul
^ permalink raw reply [flat|nested] 16+ messages in thread
* RE: [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool
2025-06-30 16:22 ` Paul Menzel
@ 2025-07-03 20:21 ` Hay, Joshua A
2025-07-07 14:43 ` Brian Vazquez
1 sibling, 0 replies; 16+ messages in thread
From: Hay, Joshua A @ 2025-07-03 20:21 UTC (permalink / raw)
To: Paul Menzel; +Cc: intel-wired-lan@lists.osuosl.org, netdev@vger.kernel.org
> >> Am 25.06.25 um 18:11 schrieb Joshua Hay:
> >>> This series fixes a stability issue in the flow scheduling Tx send/clean
> >>> path that results in a Tx timeout.
> >>>
> >>> The existing guardrails in the Tx path were not sufficient to prevent
> >>> the driver from reusing completion tags that were still in flight (held
> >>> by the HW). This collision would cause the driver to erroneously clean
> >>> the wrong packet thus leaving the descriptor ring in a bad state.
> >>>
> >>> The main point of this refactor is replace the flow scheduling buffer
> >>
> >> … to replace …?
> >
> > Thanks, will fix in v2
> >
> >>> ring with a large pool/array of buffers. The completion tag then simply
> >>> is the index into this array. The driver tracks the free tags and pulls
> >>> the next free one from a refillq. The cleaning routines simply use the
> >>> completion tag from the completion descriptor to index into the array to
> >>> quickly find the buffers to clean.
> >>>
> >>> All of the code to support the refactor is added first to ensure traffic
> >>> still passes with each patch. The final patch then removes all of the
> >>> obsolete stashing code.
> >>
> >> Do you have reproducers for the issue?
> >
> > This issue cannot be reproduced without the customer specific device
> > configuration, but it can impact any traffic once in place.
>
> Interesting. Then it’d be great if you could describe that setup in more
> detail.
Hi Paul, I'm still working with the customer on what details can be provided. I will add what I can to the cover letter in v2.
>
> >>> Joshua Hay (5):
> >>> idpf: add support for Tx refillqs in flow scheduling mode
> >>> idpf: improve when to set RE bit logic
> >>> idpf: replace flow scheduling buffer ring with buffer pool
> >>> idpf: stop Tx if there are insufficient buffer resources
> >>> idpf: remove obsolete stashing code
> >>>
> >>> .../ethernet/intel/idpf/idpf_singleq_txrx.c | 6 +-
> >>> drivers/net/ethernet/intel/idpf/idpf_txrx.c | 626 ++++++------------
> >>> drivers/net/ethernet/intel/idpf/idpf_txrx.h | 76 +--
> >>> 3 files changed, 239 insertions(+), 469 deletions(-)
>
>
> Kind regards,
>
> Paul
Thanks,
Josh
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool
2025-06-30 16:22 ` Paul Menzel
2025-07-03 20:21 ` Hay, Joshua A
@ 2025-07-07 14:43 ` Brian Vazquez
2025-07-11 21:14 ` Paul Menzel
1 sibling, 1 reply; 16+ messages in thread
From: Brian Vazquez @ 2025-07-07 14:43 UTC (permalink / raw)
To: Paul Menzel; +Cc: Joshua A Hay, intel-wired-lan, netdev, Brian Vazquez
O Mon, Jun 30, 2025 at 06:22:11PM +0200, Paul Menzel wrote:
> Dear Josh,
>
>
> Am 30.06.25 um 18:08 schrieb Hay, Joshua A:
>
> > > Am 25.06.25 um 18:11 schrieb Joshua Hay:
> > > > This series fixes a stability issue in the flow scheduling Tx send/clean
> > > > path that results in a Tx timeout.
> > > >
> > > > The existing guardrails in the Tx path were not sufficient to prevent
> > > > the driver from reusing completion tags that were still in flight (held
> > > > by the HW). This collision would cause the driver to erroneously clean
> > > > the wrong packet thus leaving the descriptor ring in a bad state.
> > > >
> > > > The main point of this refactor is replace the flow scheduling buffer
> > >
> > > … to replace …?
> >
> > Thanks, will fix in v2
> >
> > > > ring with a large pool/array of buffers. The completion tag then simply
> > > > is the index into this array. The driver tracks the free tags and pulls
> > > > the next free one from a refillq. The cleaning routines simply use the
> > > > completion tag from the completion descriptor to index into the array to
> > > > quickly find the buffers to clean.
> > > >
> > > > All of the code to support the refactor is added first to ensure traffic
> > > > still passes with each patch. The final patch then removes all of the
> > > > obsolete stashing code.
> > >
> > > Do you have reproducers for the issue?
> >
> > This issue cannot be reproduced without the customer specific device
> > configuration, but it can impact any traffic once in place.
>
> Interesting. Then it’d be great if you could describe that setup in more
> detail.
>
Hey Paul,
The hardware can process packets and return completions out of order;
this depends on HW configuration that is difficult to replicate.
To match completions with packets, each packet with pending completions
must be associated to a unique ID. The previous code would occasionally
reassigned the same ID to multiple pending packets, resulting in
resource leaks and eventually panics.
The new code uses a much simpler data structure to assign IDs that is immune to duplicate assignment, and also much more efficient at runtime.
> > > > Joshua Hay (5):
> > > > idpf: add support for Tx refillqs in flow scheduling mode
> > > > idpf: improve when to set RE bit logic
> > > > idpf: replace flow scheduling buffer ring with buffer pool
> > > > idpf: stop Tx if there are insufficient buffer resources
> > > > idpf: remove obsolete stashing code
> > > >
> > > > .../ethernet/intel/idpf/idpf_singleq_txrx.c | 6 +-
> > > > drivers/net/ethernet/intel/idpf/idpf_txrx.c | 626 ++++++------------
> > > > drivers/net/ethernet/intel/idpf/idpf_txrx.h | 76 +--
> > > > 3 files changed, 239 insertions(+), 469 deletions(-)
>
>
> Kind regards,
>
> Paul
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [Intel-wired-lan] [PATCH net 0/5] idpf: replace Tx flow scheduling buffer ring with buffer pool
2025-07-07 14:43 ` Brian Vazquez
@ 2025-07-11 21:14 ` Paul Menzel
0 siblings, 0 replies; 16+ messages in thread
From: Paul Menzel @ 2025-07-11 21:14 UTC (permalink / raw)
To: Brian Vazquez; +Cc: Joshua A Hay, intel-wired-lan, netdev
Dear Brian,
Thank you for your reply.
Am 07.07.25 um 16:43 schrieb Brian Vazquez:
> O Mon, Jun 30, 2025 at 06:22:11PM +0200, Paul Menzel wrote:
>> Am 30.06.25 um 18:08 schrieb Hay, Joshua A:
>>
>>>> Am 25.06.25 um 18:11 schrieb Joshua Hay:
>>>>> This series fixes a stability issue in the flow scheduling Tx send/clean
>>>>> path that results in a Tx timeout.
>>>>>
>>>>> The existing guardrails in the Tx path were not sufficient to prevent
>>>>> the driver from reusing completion tags that were still in flight (held
>>>>> by the HW). This collision would cause the driver to erroneously clean
>>>>> the wrong packet thus leaving the descriptor ring in a bad state.
>>>>>
>>>>> The main point of this refactor is replace the flow scheduling buffer
>>>>
>>>> … to replace …?
>>>
>>> Thanks, will fix in v2
>>>
>>>>> ring with a large pool/array of buffers. The completion tag then simply
>>>>> is the index into this array. The driver tracks the free tags and pulls
>>>>> the next free one from a refillq. The cleaning routines simply use the
>>>>> completion tag from the completion descriptor to index into the array to
>>>>> quickly find the buffers to clean.
>>>>>
>>>>> All of the code to support the refactor is added first to ensure traffic
>>>>> still passes with each patch. The final patch then removes all of the
>>>>> obsolete stashing code.
>>>>
>>>> Do you have reproducers for the issue?
>>>
>>> This issue cannot be reproduced without the customer specific device
>>> configuration, but it can impact any traffic once in place.
>>
>> Interesting. Then it’d be great if you could describe that setup in more
>> detail.
> The hardware can process packets and return completions out of order;
> this depends on HW configuration that is difficult to replicate.
>
> To match completions with packets, each packet with pending completions
> must be associated to a unique ID. The previous code would occasionally
> reassigned the same ID to multiple pending packets, resulting in
> resource leaks and eventually panics.
Thank you for describing the problem again. Too bad it’s not easily
reproducible.
> The new code uses a much simpler data structure to assign IDs that
> is immune to duplicate assignment, and also much more efficient at
> runtime.
Maybe that could be added to the commit message too. How can the
efficiency claim be verified?
>>>>> Joshua Hay (5):
>>>>> idpf: add support for Tx refillqs in flow scheduling mode
>>>>> idpf: improve when to set RE bit logic
>>>>> idpf: replace flow scheduling buffer ring with buffer pool
>>>>> idpf: stop Tx if there are insufficient buffer resources
>>>>> idpf: remove obsolete stashing code
>>>>>
>>>>> .../ethernet/intel/idpf/idpf_singleq_txrx.c | 6 +-
>>>>> drivers/net/ethernet/intel/idpf/idpf_txrx.c | 626 ++++++------------
>>>>> drivers/net/ethernet/intel/idpf/idpf_txrx.h | 76 +--
>>>>> 3 files changed, 239 insertions(+), 469 deletions(-)
Kind regards,
Paul
^ permalink raw reply [flat|nested] 16+ messages in thread