Commit b22f21f7 authored by Pavan Chebbi's avatar Pavan Chebbi Committed by David S. Miller

tg3: Improve PTP TX timestamping logic

When we are trying to timestamp a TX packet, there may be
occasions when the TX timestamp register is still not
updated with the latest timestamp even if the timestamp
packet descriptor is marked as complete.
This usually happens in cases where the system is under
stress or flow control is affecting the transmit side.

We will solve this problem by saving the snapshot of the
timestamp register when we are posting the TX descriptor.
At this time, the register contains previously timestamped
packet's value and valid timestamp of the current packet must
be different than this.
Upon completion of the current descriptor, we will check if
the timestamp register is updated or not before timestamping
the skb. If not updated, we will schedule the ptp worker to
fetch the updated time later and timestamp the skb.
Also now we restrict number of outstanding PTP TX packet
requests to 1.
Reported-by: default avatarSimon White <Simon.White@viavisolutions.com>
Link: https://lore.kernel.org/netdev/CACKFLikGdN9XPtWk-fdrzxdcD=+bv-GHBvfVfSpJzHY7hrW39g@mail.gmail.com/Signed-off-by: default avatarPavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: default avatarMichael Chan <michael.chan@broadcom.com>
Reviewed-by: default avatarAndy Gospodarek <gospo@broadcom.com>
Reviewed-by: default avatarKalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 6e55b1cb
......@@ -6314,6 +6314,46 @@ static int tg3_ptp_enable(struct ptp_clock_info *ptp,
return -EOPNOTSUPP;
}
static void tg3_hwclock_to_timestamp(struct tg3 *tp, u64 hwclock,
struct skb_shared_hwtstamps *timestamp)
{
memset(timestamp, 0, sizeof(struct skb_shared_hwtstamps));
timestamp->hwtstamp = ns_to_ktime((hwclock & TG3_TSTAMP_MASK) +
tp->ptp_adjust);
}
static void tg3_read_tx_tstamp(struct tg3 *tp, u64 *hwclock)
{
*hwclock = tr32(TG3_TX_TSTAMP_LSB);
*hwclock |= (u64)tr32(TG3_TX_TSTAMP_MSB) << 32;
}
static long tg3_ptp_ts_aux_work(struct ptp_clock_info *ptp)
{
struct tg3 *tp = container_of(ptp, struct tg3, ptp_info);
struct skb_shared_hwtstamps timestamp;
u64 hwclock;
if (tp->ptp_txts_retrycnt > 2)
goto done;
tg3_read_tx_tstamp(tp, &hwclock);
if (hwclock != tp->pre_tx_ts) {
tg3_hwclock_to_timestamp(tp, hwclock, &timestamp);
skb_tstamp_tx(tp->tx_tstamp_skb, &timestamp);
goto done;
}
tp->ptp_txts_retrycnt++;
return HZ / 10;
done:
dev_consume_skb_any(tp->tx_tstamp_skb);
tp->tx_tstamp_skb = NULL;
tp->ptp_txts_retrycnt = 0;
tp->pre_tx_ts = 0;
return -1;
}
static const struct ptp_clock_info tg3_ptp_caps = {
.owner = THIS_MODULE,
.name = "tg3 clock",
......@@ -6325,19 +6365,12 @@ static const struct ptp_clock_info tg3_ptp_caps = {
.pps = 0,
.adjfine = tg3_ptp_adjfine,
.adjtime = tg3_ptp_adjtime,
.do_aux_work = tg3_ptp_ts_aux_work,
.gettimex64 = tg3_ptp_gettimex,
.settime64 = tg3_ptp_settime,
.enable = tg3_ptp_enable,
};
static void tg3_hwclock_to_timestamp(struct tg3 *tp, u64 hwclock,
struct skb_shared_hwtstamps *timestamp)
{
memset(timestamp, 0, sizeof(struct skb_shared_hwtstamps));
timestamp->hwtstamp = ns_to_ktime((hwclock & TG3_TSTAMP_MASK) +
tp->ptp_adjust);
}
/* tp->lock must be held */
static void tg3_ptp_init(struct tg3 *tp)
{
......@@ -6368,6 +6401,8 @@ static void tg3_ptp_fini(struct tg3 *tp)
ptp_clock_unregister(tp->ptp_clock);
tp->ptp_clock = NULL;
tp->ptp_adjust = 0;
dev_consume_skb_any(tp->tx_tstamp_skb);
tp->tx_tstamp_skb = NULL;
}
static inline int tg3_irq_sync(struct tg3 *tp)
......@@ -6538,6 +6573,7 @@ static void tg3_tx(struct tg3_napi *tnapi)
while (sw_idx != hw_idx) {
struct tg3_tx_ring_info *ri = &tnapi->tx_buffers[sw_idx];
bool complete_skb_later = false;
struct sk_buff *skb = ri->skb;
int i, tx_bug = 0;
......@@ -6548,12 +6584,17 @@ static void tg3_tx(struct tg3_napi *tnapi)
if (tnapi->tx_ring[sw_idx].len_flags & TXD_FLAG_HWTSTAMP) {
struct skb_shared_hwtstamps timestamp;
u64 hwclock = tr32(TG3_TX_TSTAMP_LSB);
hwclock |= (u64)tr32(TG3_TX_TSTAMP_MSB) << 32;
tg3_hwclock_to_timestamp(tp, hwclock, &timestamp);
u64 hwclock;
skb_tstamp_tx(skb, &timestamp);
tg3_read_tx_tstamp(tp, &hwclock);
if (hwclock != tp->pre_tx_ts) {
tg3_hwclock_to_timestamp(tp, hwclock, &timestamp);
skb_tstamp_tx(skb, &timestamp);
tp->pre_tx_ts = 0;
} else {
tp->tx_tstamp_skb = skb;
complete_skb_later = true;
}
}
dma_unmap_single(&tp->pdev->dev, dma_unmap_addr(ri, mapping),
......@@ -6591,7 +6632,10 @@ static void tg3_tx(struct tg3_napi *tnapi)
pkts_compl++;
bytes_compl += skb->len;
dev_consume_skb_any(skb);
if (!complete_skb_later)
dev_consume_skb_any(skb);
else
ptp_schedule_worker(tp->ptp_clock, 0);
if (unlikely(tx_bug)) {
tg3_tx_recover(tp);
......@@ -8028,8 +8072,13 @@ static netdev_tx_t tg3_start_xmit(struct sk_buff *skb, struct net_device *dev)
if ((unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) &&
tg3_flag(tp, TX_TSTAMP_EN)) {
skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
base_flags |= TXD_FLAG_HWTSTAMP;
tg3_full_lock(tp, 0);
if (!tp->pre_tx_ts) {
skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
base_flags |= TXD_FLAG_HWTSTAMP;
tg3_read_tx_tstamp(tp, &tp->pre_tx_ts);
}
tg3_full_unlock(tp);
}
len = skb_headlen(skb);
......
......@@ -3190,6 +3190,7 @@ struct tg3 {
struct ptp_clock_info ptp_info;
struct ptp_clock *ptp_clock;
s64 ptp_adjust;
u8 ptp_txts_retrycnt;
/* begin "tx thread" cacheline section */
void (*write32_tx_mbox) (struct tg3 *, u32,
......@@ -3372,6 +3373,8 @@ struct tg3 {
struct tg3_hw_stats *hw_stats;
dma_addr_t stats_mapping;
struct work_struct reset_task;
struct sk_buff *tx_tstamp_skb;
u64 pre_tx_ts;
int nvram_lock_cnt;
u32 nvram_size;
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment