* [PATCH 0/2] brcmfmac: Stability and throughput changes
@ 2020-06-22 2:07 Wright Feng
2020-06-22 2:07 ` [PATCH 1/2] brcmfmac: set state of hanger slot to FREE when flushing PSQ Wright Feng
2020-06-22 2:07 ` [PATCH 2/2] brcmfmac: Set pacing shift before transmitting skb to bus Wright Feng
0 siblings, 2 replies; 6+ messages in thread
From: Wright Feng @ 2020-06-22 2:07 UTC (permalink / raw)
To: linux-wireless
Cc: wright.feng, brcm80211-dev-list, brcm80211-dev-list,
Arend van Spriel, Franky Lin, Hante Meuleman, Kalle Valo,
chi-hsien.lin
This series fixes stability issue and enhances throughput.
Wright Feng (2):
brcmfmac: set state of hanger slot to FREE when flushing PSQ
brcmfmac: Set pacing shift before transmitting skb to bus
drivers/net/wireless/broadcom/brcm80211/brcmfmac/core.c | 3 +++
drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c | 4 ++++
2 files changed, 7 insertions(+)
--
2.25.0
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH 1/2] brcmfmac: set state of hanger slot to FREE when flushing PSQ
2020-06-22 2:07 [PATCH 0/2] brcmfmac: Stability and throughput changes Wright Feng
@ 2020-06-22 2:07 ` Wright Feng
2020-06-22 8:58 ` Arend Van Spriel
2020-06-22 2:07 ` [PATCH 2/2] brcmfmac: Set pacing shift before transmitting skb to bus Wright Feng
1 sibling, 1 reply; 6+ messages in thread
From: Wright Feng @ 2020-06-22 2:07 UTC (permalink / raw)
To: linux-wireless
Cc: wright.feng, brcm80211-dev-list, brcm80211-dev-list,
Arend van Spriel, Franky Lin, Hante Meuleman, Kalle Valo,
chi-hsien.lin
When USB or SDIO device got abnormal bus disconnection, host driver
tried to clean up the skbs in PSQ and TXQ (The skb's pointer in hanger
slot linked to PSQ and TSQ), so we should set the state of skb hanger slot
to BRCMF_FWS_HANGER_ITEM_STATE_FREE before freeing skb.
In brcmf_fws_bus_txq_cleanup it already sets
BRCMF_FWS_HANGER_ITEM_STATE_FREE before freeing skb, therefore we add the
same thing in brcmf_fws_psq_flush to avoid following warning message.
[ 1580.012880] ------------ [ cut here ]------------
[ 1580.017550] WARNING: CPU: 3 PID: 3065 at
drivers/net/wireless/broadcom/brcm80211/brcmutil/utils.c:49
brcmu_pkt_buf_free_skb+0x21/0x30 [brcmutil]
[ 1580.184017] Call Trace:
[ 1580.186514] brcmf_fws_cleanup+0x14e/0x190 [brcmfmac]
[ 1580.191594] brcmf_fws_del_interface+0x70/0x90 [brcmfmac]
[ 1580.197029] brcmf_proto_bcdc_del_if+0xe/0x10 [brcmfmac]
[ 1580.202418] brcmf_remove_interface+0x69/0x190 [brcmfmac]
[ 1580.207888] brcmf_detach+0x90/0xe0 [brcmfmac]
[ 1580.212385] brcmf_usb_disconnect+0x76/0xb0 [brcmfmac]
[ 1580.217557] usb_unbind_interface+0x72/0x260
[ 1580.221857] device_release_driver_internal+0x141/0x200
[ 1580.227152] device_release_driver+0x12/0x20
[ 1580.231460] bus_remove_device+0xfd/0x170
[ 1580.235504] device_del+0x1d9/0x300
[ 1580.239041] usb_disable_device+0x9e/0x270
[ 1580.243160] usb_disconnect+0x94/0x270
[ 1580.246980] hub_event+0x76d/0x13b0
[ 1580.250499] process_one_work+0x144/0x360
[ 1580.254564] worker_thread+0x4d/0x3c0
[ 1580.258247] kthread+0x109/0x140
[ 1580.261515] ? rescuer_thread+0x340/0x340
[ 1580.265543] ? kthread_park+0x60/0x60
[ 1580.269237] ? SyS_exit_group+0x14/0x20
[ 1580.273118] ret_from_fork+0x25/0x30
[ 1580.300446] ------------ [ cut here ]------------
Signed-off-by: Wright Feng <wright.feng@cypress.com>
Signed-off-by: Chi-hsien Lin <chi-hsien.lin@cypress.com>
---
drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c
index 09701262330d..babaac682f13 100644
--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c
@@ -621,6 +621,7 @@ static inline int brcmf_fws_hanger_poppkt(struct brcmf_fws_hanger *h,
static void brcmf_fws_psq_flush(struct brcmf_fws_info *fws, struct pktq *q,
int ifidx)
{
+ struct brcmf_fws_hanger_item *hi;
bool (*matchfn)(struct sk_buff *, void *) = NULL;
struct sk_buff *skb;
int prec;
@@ -632,6 +633,9 @@ static void brcmf_fws_psq_flush(struct brcmf_fws_info *fws, struct pktq *q,
skb = brcmu_pktq_pdeq_match(q, prec, matchfn, &ifidx);
while (skb) {
hslot = brcmf_skb_htod_tag_get_field(skb, HSLOT);
+ hi = &fws->hanger.items[hslot];
+ WARN_ON(skb != hi->pkt);
+ hi->state = BRCMF_FWS_HANGER_ITEM_STATE_FREE;
brcmf_fws_hanger_poppkt(&fws->hanger, hslot, &skb,
true);
brcmu_pkt_buf_free_skb(skb);
--
2.25.0
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH 2/2] brcmfmac: Set pacing shift before transmitting skb to bus
2020-06-22 2:07 [PATCH 0/2] brcmfmac: Stability and throughput changes Wright Feng
2020-06-22 2:07 ` [PATCH 1/2] brcmfmac: set state of hanger slot to FREE when flushing PSQ Wright Feng
@ 2020-06-22 2:07 ` Wright Feng
2020-06-22 8:57 ` Arend Van Spriel
1 sibling, 1 reply; 6+ messages in thread
From: Wright Feng @ 2020-06-22 2:07 UTC (permalink / raw)
To: linux-wireless
Cc: wright.feng, brcm80211-dev-list, brcm80211-dev-list,
Arend van Spriel, Franky Lin, Hante Meuleman, Kalle Valo,
chi-hsien.lin
Linux 3.6 introduces TSQ which has a per socket threshold for TCP Tx
packet to reduce latency. In flow control mode, host driver enqueues skb
in hanger and TCP doesn't push new skb frees until host frees the skb when
receiving fwstatus event. So set pacing shift 8 to send them as a single
large aggregate frame to the bus layer.
43455 TX TCP throughput in FC mode 2 on Linux 5.4.18
sk_pacing_shift : Throughput
10: 190 Mbps
9: 216 Mbps
8: 233 Mbps (Max throughput)
7: 233 Mpbs
Signed-off-by: Wright Feng <wright.feng@cypress.com>
Signed-off-by: Chi-hsien Lin <chi-hsien.lin@cypress.com>
---
drivers/net/wireless/broadcom/brcm80211/brcmfmac/core.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/core.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/core.c
index c88655acc78c..f89010a81ffb 100644
--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/core.c
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/core.c
@@ -352,6 +352,9 @@ static netdev_tx_t brcmf_netdev_start_xmit(struct sk_buff *skb,
if ((skb->priority == 0) || (skb->priority > 7))
skb->priority = cfg80211_classify8021d(skb, NULL);
+ /* set pacing shift for packet aggregation */
+ sk_pacing_shift_update(skb->sk, 8);
+
ret = brcmf_proto_tx_queue_data(drvr, ifp->ifidx, skb);
if (ret < 0)
brcmf_txfinalize(ifp, skb, false);
--
2.25.0
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH 2/2] brcmfmac: Set pacing shift before transmitting skb to bus
2020-06-22 2:07 ` [PATCH 2/2] brcmfmac: Set pacing shift before transmitting skb to bus Wright Feng
@ 2020-06-22 8:57 ` Arend Van Spriel
2020-06-24 5:45 ` Wright Feng
0 siblings, 1 reply; 6+ messages in thread
From: Arend Van Spriel @ 2020-06-22 8:57 UTC (permalink / raw)
To: Wright Feng, linux-wireless
Cc: brcm80211-dev-list, brcm80211-dev-list, Franky Lin,
Hante Meuleman, Kalle Valo, chi-hsien.lin
On 6/22/2020 4:07 AM, Wright Feng wrote:
> Linux 3.6 introduces TSQ which has a per socket threshold for TCP Tx
> packet to reduce latency. In flow control mode, host driver enqueues skb
> in hanger and TCP doesn't push new skb frees until host frees the skb when
> receiving fwstatus event. So set pacing shift 8 to send them as a single
> large aggregate frame to the bus layer.
>
> 43455 TX TCP throughput in FC mode 2 on Linux 5.4.18
> sk_pacing_shift : Throughput
> 10: 190 Mbps
> 9: 216 Mbps
> 8: 233 Mbps (Max throughput)
> 7: 233 Mpbs
In this patch the pacing shift update is done irrespective of the FC
mode so I would like to see similar measurements for the other FC mode
configurations.
Regards,
Arend
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH 1/2] brcmfmac: set state of hanger slot to FREE when flushing PSQ
2020-06-22 2:07 ` [PATCH 1/2] brcmfmac: set state of hanger slot to FREE when flushing PSQ Wright Feng
@ 2020-06-22 8:58 ` Arend Van Spriel
0 siblings, 0 replies; 6+ messages in thread
From: Arend Van Spriel @ 2020-06-22 8:58 UTC (permalink / raw)
To: Wright Feng, linux-wireless
Cc: brcm80211-dev-list, brcm80211-dev-list, Franky Lin,
Hante Meuleman, Kalle Valo, chi-hsien.lin
On 6/22/2020 4:07 AM, Wright Feng wrote:
> When USB or SDIO device got abnormal bus disconnection, host driver
> tried to clean up the skbs in PSQ and TXQ (The skb's pointer in hanger
> slot linked to PSQ and TSQ), so we should set the state of skb hanger slot
> to BRCMF_FWS_HANGER_ITEM_STATE_FREE before freeing skb.
> In brcmf_fws_bus_txq_cleanup it already sets
> BRCMF_FWS_HANGER_ITEM_STATE_FREE before freeing skb, therefore we add the
> same thing in brcmf_fws_psq_flush to avoid following warning message.
[...]
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
> Signed-off-by: Wright Feng <wright.feng@cypress.com>
> Signed-off-by: Chi-hsien Lin <chi-hsien.lin@cypress.com>
> ---
> drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c | 4 ++++
> 1 file changed, 4 insertions(+)
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH 2/2] brcmfmac: Set pacing shift before transmitting skb to bus
2020-06-22 8:57 ` Arend Van Spriel
@ 2020-06-24 5:45 ` Wright Feng
0 siblings, 0 replies; 6+ messages in thread
From: Wright Feng @ 2020-06-24 5:45 UTC (permalink / raw)
To: Arend Van Spriel, linux-wireless
Cc: brcm80211-dev-list, brcm80211-dev-list, Franky Lin,
Hante Meuleman, Kalle Valo, chi-hsien.lin
Arend Van Spriel 於 6/22/2020 4:57 PM 寫道:
> On 6/22/2020 4:07 AM, Wright Feng wrote:
>> Linux 3.6 introduces TSQ which has a per socket threshold for TCP Tx
>> packet to reduce latency. In flow control mode, host driver enqueues skb
>> in hanger and TCP doesn't push new skb frees until host frees the skb
>> when
>> receiving fwstatus event. So set pacing shift 8 to send them as a single
>> large aggregate frame to the bus layer.
>>
>> 43455 TX TCP throughput in FC mode 2 on Linux 5.4.18
>> sk_pacing_shift : Throughput
>> 10: 190 Mbps
>> 9: 216 Mbps
>> 8: 233 Mbps (Max throughput)
>> 7: 233 Mpbs
>
> In this patch the pacing shift update is done irrespective of the FC
> mode so I would like to see similar measurements for the other FC mode
> configurations.
>
Measure 43455 TX TCP throughput again with different FC mode.
Here is the result and I will update result in patch v2.
43455 TX TCP throughput on Linux 5.4.18
FC mode 0
sk_pacing_shift : Throughput
10: 245 Mbps
9: 245 Mbps
8: 246 Mbps
7: 246 Mbps
FC mode 1
sk_pacing_shift : Throughput
10: 182 Mbps
9: 197 Mbps
8: 206 Mbps
7: 207 Mbps
FC mode 2
sk_pacing_shift : Throughput
10: 180 Mbps
9: 197 Mbps
8: 206 Mbps
7: 207 Mbps
> Regards,
> Arend
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2020-06-24 5:45 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-06-22 2:07 [PATCH 0/2] brcmfmac: Stability and throughput changes Wright Feng
2020-06-22 2:07 ` [PATCH 1/2] brcmfmac: set state of hanger slot to FREE when flushing PSQ Wright Feng
2020-06-22 8:58 ` Arend Van Spriel
2020-06-22 2:07 ` [PATCH 2/2] brcmfmac: Set pacing shift before transmitting skb to bus Wright Feng
2020-06-22 8:57 ` Arend Van Spriel
2020-06-24 5:45 ` Wright Feng
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).