Skip to contentSkip to conversation

r8169: fix jumbo packet handling on resume from suspend

PATCH In review Heiner Kallweit 2 participants 4 messages Oct 9, 2019 – Oct 12, 2019

Subsystems (3)
lore
Share & tools
lore

Heiner Kallweit

author maintainer
#0 lore raw

PATCH 1/1 r8169: fix jumbo packet handling on resume from suspend

Mariusz reported that invalid packets are sent after resume from suspend if jumbo packets are active. It turned out that his BIOS resets chip settings to non-jumbo on resume. Most chip settings are re-initialized on resume from suspend by calling rtl_hw_start(), so let's add configuring jumbo to this function.
There's nothing wrong with the commit marked as fixed, it's just the first one where the patch applies cleanly.

Fixes 7366016d2d4c ("r8169: read common register for PCI commit")
---
 drivers/net/ethernet/realtek/r8169_main.c | 35 +++++++----------------
 1 file changed, 11 insertions(+), 24 deletions(-)
drivers/net/ethernet/realtek/r8169_main.c +11 -24
index 74f81fe03..350b0d949 100644
@@ -4146,6 +4146,14 @@ static void rtl_hw_jumbo_disable(struct rtl8169_private *tp)
 	rtl_lock_config_regs(tp);
 }
 
+static void rtl_jumbo_config(struct rtl8169_private *tp, int mtu)
+{
+	if (mtu > ETH_DATA_LEN)
+		rtl_hw_jumbo_enable(tp);
+	else
+		rtl_hw_jumbo_disable(tp);
+}
+
 DECLARE_RTL_COND(rtl_chipcmd_cond)
 {
 	return RTL_R8(tp, ChipCmd) & CmdReset;
@@ -4442,11 +4450,6 @@ static void rtl8168g_set_pause_thresholds(struct rtl8169_private *tp,
 static void rtl_hw_start_8168bb(struct rtl8169_private *tp)
 {
 	RTL_W8(tp, Config3, RTL_R8(tp, Config3) & ~Beacon_en);
-
-	if (tp->dev->mtu <= ETH_DATA_LEN) {
-		rtl_tx_performance_tweak(tp, PCI_EXP_DEVCTL_READRQ_4096B |
-					 PCI_EXP_DEVCTL_NOSNOOP_EN);
-	}
 }
 
 static void rtl_hw_start_8168bef(struct rtl8169_private *tp)
@@ -4462,9 +4465,6 @@ static void __rtl_hw_start_8168cp(struct rtl8169_private *tp)
 
 	RTL_W8(tp, Config3, RTL_R8(tp, Config3) & ~Beacon_en);
 
-	if (tp->dev->mtu <= ETH_DATA_LEN)
-		rtl_tx_performance_tweak(tp, PCI_EXP_DEVCTL_READRQ_4096B);
-
 	rtl_disable_clock_request(tp);
 }
 
@@ -4490,9 +4490,6 @@ static void rtl_hw_start_8168cp_2(struct rtl8169_private *tp)
 	rtl_set_def_aspm_entry_latency(tp);
 
 	RTL_W8(tp, Config3, RTL_R8(tp, Config3) & ~Beacon_en);
-
-	if (tp->dev->mtu <= ETH_DATA_LEN)
-		rtl_tx_performance_tweak(tp, PCI_EXP_DEVCTL_READRQ_4096B);
 }
 
 static void rtl_hw_start_8168cp_3(struct rtl8169_private *tp)
@@ -4503,9 +4500,6 @@ static void rtl_hw_start_8168cp_3(struct rtl8169_private *tp)
 
 	/* Magic. */
 	RTL_W8(tp, DBG_REG, 0x20);
-
-	if (tp->dev->mtu <= ETH_DATA_LEN)
-		rtl_tx_performance_tweak(tp, PCI_EXP_DEVCTL_READRQ_4096B);
 }
 
 static void rtl_hw_start_8168c_1(struct rtl8169_private *tp)
@@ -4611,9 +4605,6 @@ static void rtl_hw_start_8168e_1(struct rtl8169_private *tp)
 
 	rtl_ephy_init(tp, e_info_8168e_1);
 
-	if (tp->dev->mtu <= ETH_DATA_LEN)
-		rtl_tx_performance_tweak(tp, PCI_EXP_DEVCTL_READRQ_4096B);
-
 	rtl_disable_clock_request(tp);
 
 	/* Reset tx FIFO pointer */
@@ -4636,9 +4627,6 @@ static void rtl_hw_start_8168e_2(struct rtl8169_private *tp)
 
 	rtl_ephy_init(tp, e_info_8168e_2);
 
-	if (tp->dev->mtu <= ETH_DATA_LEN)
-		rtl_tx_performance_tweak(tp, PCI_EXP_DEVCTL_READRQ_4096B);
-
 	rtl_eri_write(tp, 0xc0, ERIAR_MASK_0011, 0x0000);
 	rtl_eri_write(tp, 0xb8, ERIAR_MASK_0011, 0x0000);
 	rtl_set_fifo_size(tp, 0x10, 0x10, 0x02, 0x06);
@@ -5485,6 +5473,8 @@ static void rtl_hw_start(struct  rtl8169_private *tp)
 	rtl_set_rx_tx_desc_registers(tp);
 	rtl_lock_config_regs(tp);
 
+	rtl_jumbo_config(tp, tp->dev->mtu);
+
 	/* Initially a 10 us delay. Turned it into a PCI commit. - FR */
 	RTL_R16(tp, CPlusCmd);
 	RTL_W8(tp, ChipCmd, CmdTxEnb | CmdRxEnb);
@@ -5498,10 +5488,7 @@ static int rtl8169_change_mtu(struct net_device *dev, int new_mtu)
 {
 	struct rtl8169_private *tp = netdev_priv(dev);
 
-	if (new_mtu > ETH_DATA_LEN)
-		rtl_hw_jumbo_enable(tp);
-	else
-		rtl_hw_jumbo_disable(tp);
+	rtl_jumbo_config(tp, new_mtu);
 
 	dev->mtu = new_mtu;
 	netdev_update_features(dev);
Signature · 1 line
2.23.0
#1 lore raw
Quoted 8 lines · Heiner Kallweit

Mariusz reported that invalid packets are sent after resume from suspend if jumbo packets are active. It turned out that his BIOS resets chip settings to non-jumbo on resume. Most chip settings are re-initialized on resume from suspend by calling rtl_hw_start(), so let's add configuring jumbo to this function.
There's nothing wrong with the commit marked as fixed, it's just the first one where the patch applies cleanly.

Quoted 4 lines
Fixes 7366016d2d4c ("r8169: read common register for PCI commit")

Applied, somewhat begrudgingly - this really isn't the way the Fixes tag should be used, but I appreciate it may be hard at this point to pin down a commit to blame given how many generations of HW this driver supports and how old it is.. perhaps I should have removed the tag in this case, hm.

Since the selected commit came in 5.4 I'm not queuing for stable.

author maintainer
#2 lore raw
Quoted 14 lines · Jakub Kicinski
Quoted 8 lines · Heiner Kallweit

Mariusz reported that invalid packets are sent after resume from suspend if jumbo packets are active. It turned out that his BIOS resets chip settings to non-jumbo on resume. Most chip settings are re-initialized on resume from suspend by calling rtl_hw_start(), so let's add configuring jumbo to this function.
There's nothing wrong with the commit marked as fixed, it's just the first one where the patch applies cleanly.

Quoted 4 lines
Fixes 7366016d2d4c ("r8169: read common register for PCI commit")
Quoted 8 lines

Applied, somewhat begrudgingly - this really isn't the way the Fixes tag should be used, but I appreciate it may be hard at this point to pin down a commit to blame given how many generations of HW this driver supports and how old it is.. perhaps I should have removed the tag in this case, hm.

Since the selected commit came in 5.4 I'm not queuing for stable.

The issue seems to have been there forever, but patch applies from a certain kernel version only. I agree that using the Fixes tag to provide this information is kind of a misuse. How would you prefer to get that information, add a comment below the commit message similar to the list of changes in a new version of a patch series?

Heiner

#3 lore raw
Quoted 21 lines · Heiner Kallweit
Quoted 20 lines · Jakub Kicinski
Quoted 8 lines · Heiner Kallweit

Mariusz reported that invalid packets are sent after resume from suspend if jumbo packets are active. It turned out that his BIOS resets chip settings to non-jumbo on resume. Most chip settings are re-initialized on resume from suspend by calling rtl_hw_start(), so let's add configuring jumbo to this function.
There's nothing wrong with the commit marked as fixed, it's just the first one where the patch applies cleanly.

Quoted 4 lines
Fixes 7366016d2d4c ("r8169: read common register for PCI commit")

Applied, somewhat begrudgingly - this really isn't the way the Fixes tag should be used, but I appreciate it may be hard at this point to pin down a commit to blame given how many generations of HW this driver supports and how old it is.. perhaps I should have removed the tag in this case, hm.

Quoted 7 lines
Quoted 2 lines

Since the selected commit came in 5.4 I'm not queuing for stable.

The issue seems to have been there forever, but patch applies from a certain kernel version only. I agree that using the Fixes tag to provide this information is kind of a misuse. How would you prefer to get that information, add a comment below the commit message similar to the list of changes in a new version of a patch series?

I'd put the backport help under the --- lines, maybe additionally mentioning its presence in the commit message (lore link would complete the picture). Like we do for merges.

Although I think Dave queues for stable immediately when patch is merged to net, so if the backport is to last release or two I think the info under --- could be as useful as in the commit message.

Another way would be posting the backported patch (say for the most recent LTS) if the backport is hard and fix important 🤔