Skip to contentSkip to conversation

thunderbolt: Fixes for few reported issues

PATCH Applied Mika Westerberg 2 participants 6 messages Oct 1, 2019 – Oct 8, 2019

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover thunderbolt: Fixes for few reported issues

Hi all,

This series includes fixes for a couple of issues people have reported:

- Discovering DP path fails on Light Ridge based iMac with two devices (monitors) connected.

- There is a lockdep splat on Dominik's system when plugging in Thunderbolt dock.

- Nicholas spotted that we do unnecessary PCI config space read when issuing LC mailbox command on ICL.

Mika Westerberg (3):
thunderbolt: Read DP IN adapter first two dwords in one go
thunderbolt: Fix lockdep circular locking depedency warning
thunderbolt: Drop unnecessary read when writing LC command in Ice Lake

 drivers/thunderbolt/nhi_ops.c |  1 -
 drivers/thunderbolt/switch.c  | 28 +++++++++++-----------------
 2 files changed, 11 insertions(+), 18 deletions(-)
Signature · 1 line
2.23.0
#3 lore raw

PATCH 2/3 thunderbolt: Fix lockdep circular locking depedency warning

When lockdep is enabled, plugging Thunderbolt dock on Dominik's laptop triggers following splat:

======================================================
WARNING: possible circular locking dependency detected
5.3.0-rc6+ #1 Tainted: G T
------------------------------------------------------
pool-/usr/lib/b/1258 is trying to acquire lock:
000000005ab0ad43 (pci_rescan_remove_lock){+.+.}, at: authorized_store+0xe8/0x210

but task is already holding lock:
00000000bfb796b5 (&tb->lock){+.+.}, at: authorized_store+0x7c/0x210

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (&tb->lock){+.+.}:

         __mutex_lock+0xac/0x9a0
         tb_domain_add+0x2d/0x130
         nhi_probe+0x1dd/0x330
         pci_device_probe+0xd2/0x150
         really_probe+0xee/0x280
         driver_probe_device+0x50/0xc0
         bus_for_each_drv+0x84/0xd0
         __device_attach+0xe4/0x150
         pci_bus_add_device+0x4e/0x70
         pci_bus_add_devices+0x2e/0x66
         pci_bus_add_devices+0x59/0x66
         pci_bus_add_devices+0x59/0x66
         enable_slot+0x344/0x450
         acpiphp_check_bridge.part.0+0x119/0x150
         acpiphp_hotplug_notify+0xaa/0x140
         acpi_device_hotplug+0xa2/0x3f0
         acpi_hotplug_work_fn+0x1a/0x30
         process_one_work+0x234/0x580
         worker_thread+0x50/0x3b0
         kthread+0x10a/0x140
         ret_from_fork+0x3a/0x50

-> #0 (pci_rescan_remove_lock){+.+.}:

         __lock_acquire+0xe54/0x1ac0
         lock_acquire+0xb8/0x1b0
         __mutex_lock+0xac/0x9a0
         authorized_store+0xe8/0x210
         kernfs_fop_write+0x125/0x1b0
         vfs_write+0xc2/0x1d0
         ksys_write+0x6c/0xf0
         do_syscall_64+0x50/0x180
         entry_SYSCALL_64_after_hwframe+0x49/0xbe
  other info that might help us debug this:
   Possible unsafe locking scenario:
         CPU0                    CPU1
         ----                    ----
    lock(&tb->lock);
                                 lock(pci_rescan_remove_lock);
                                 lock(&tb->lock);
    lock(pci_rescan_remove_lock);

*** DEADLOCK ***
5 locks held by pool-/usr/lib/b/1258:
#0: 000000003df1a1ad (&f->f_pos_lock){+.+.}, at: __fdget_pos+0x4d/0x60 #1: 0000000095a40b02 (sb_writers#6){.+.+}, at: vfs_write+0x185/0x1d0 #2: 0000000017a7d714 (&of->mutex){+.+.}, at: kernfs_fop_write+0xf2/0x1b0 #3: 000000004f262981 (kn->count#208){.+.+}, at: kernfs_fop_write+0xfa/0x1b0 #4: 00000000bfb796b5 (&tb->lock){+.+.}, at: authorized_store+0x7c/0x210

  stack backtrace:
  CPU: 0 PID: 1258 Comm: pool-/usr/lib/b Tainted: G                T 5.3.0-rc6+ #1

On an system using ACPI hotplug the host router gets hotplugged first and then the firmware starts sending notifications about connected devices so the above scenario should not happen in reality. However, after taking a second look at commit a03e828915c0 ("thunderbolt: Serialize PCIe tunnel creation with PCI rescan") that introduced the locking, I don't think it is actually correct. It may have cured the symptom but probably the real root cause was somewhere closer to PCI stack and possibly is already fixed with recent kernels. I also tried to reproduce the original issue with the commit reverted but could not.

So to keep lockdep happy and the code bit less complex drop calls to pci_lock_rescan_remove()/pci_unlock_rescan_remove() in tb_switch_set_authorized() effectively reverting a03e828915c0.

Fixes a03e828915c0 ("thunderbolt: Serialize PCIe tunnel creation with PCI rescan")
---
 drivers/thunderbolt/switch.c | 9 ---------
 1 file changed, 9 deletions(-)
drivers/thunderbolt/switch.c +0 -9
index 8e712fbf8233..5ea8db667e83 100644
@@ -1034,13 +1034,6 @@ static int tb_switch_set_authorized(struct tb_switch *sw, unsigned int val)
 	if (sw->authorized)
 		goto unlock;
 
-	/*
-	 * Make sure there is no PCIe rescan ongoing when a new PCIe
-	 * tunnel is created. Otherwise the PCIe rescan code might find
-	 * the new tunnel too early.
-	 */
-	pci_lock_rescan_remove();
-
 	switch (val) {
 	/* Approve switch */
 	case 1:
@@ -1060,8 +1053,6 @@ static int tb_switch_set_authorized(struct tb_switch *sw, unsigned int val)
 		break;
 	}
 
-	pci_unlock_rescan_remove();
-
 	if (!ret) {
 		sw->authorized = val;
 		/* Notify status change to the userspace */
Signature · 1 line
2.23.0
#4 lore raw

PATCH 3/3 thunderbolt: Drop unnecessary read when writing LC command in Ice Lake

In reply to Mika Westerberg

The read is not needed as we overwrite the returned value in the next line anyway so drop it.

Fixes 3cdb9446a117 ("thunderbolt: Add support for Intel Ice Lake")
---
 drivers/thunderbolt/nhi_ops.c | 1 -
 1 file changed, 1 deletion(-)
drivers/thunderbolt/nhi_ops.c +0 -1
index 61cd09cef943..6795851aac95 100644
@@ -80,7 +80,6 @@ static void icl_nhi_lc_mailbox_cmd(struct tb_nhi *nhi, enum icl_lc_mailbox_cmd c
 {
 	u32 data;
 
-	pci_read_config_dword(nhi->pdev, VS_CAP_19, &data);
 	data = (cmd << VS_CAP_19_CMD_SHIFT) & VS_CAP_19_CMD_MASK;
 	pci_write_config_dword(nhi->pdev, VS_CAP_19, data | VS_CAP_19_VALID);
 }
Signature · 1 line
2.23.0
#1 lore raw

PATCH 1/3 thunderbolt: Read DP IN adapter first two dwords in one go

In reply to Mika Westerberg

When we discover existing DP tunnels the code checks whether DP IN adapter port is enabled by calling tb_dp_port_is_enabled() before it continues the discovery process. On Light Ridge (gen 1) controller reading only the first dword of the DP IN config space causes subsequent access to the same DP IN port path config space to fail or return invalid data as can be seen in the below splat:

thunderbolt 0000:07:00.0: CFG_ERROR(0:d): Invalid config space or offset

  Call Trace:
   tb_cfg_read+0xb9/0xd0
   __tb_path_deactivate_hop+0x98/0x210
   tb_path_activate+0x228/0x7d0
   tb_tunnel_restart+0x95/0x200
   tb_handle_hotplug+0x30e/0x630
   process_one_work+0x1b4/0x340
   worker_thread+0x44/0x3d0
   kthread+0xeb/0x120
   ? process_one_work+0x340/0x340
   ? kthread_park+0xa0/0xa0
   ret_from_fork+0x1f/0x30

If both DP In adapter config dwords are read in one go the issue does not reproduce. This is likely firmware bug but we can work it around by always reading the two dwords in one go. There should be no harm for other controllers either so can do it unconditionally.

---
 drivers/thunderbolt/switch.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)
drivers/thunderbolt/switch.c +11 -8
index 410bf1bceeee..8e712fbf8233 100644
@@ -896,12 +896,13 @@ int tb_dp_port_set_hops(struct tb_port *port, unsigned int video,
  */
 bool tb_dp_port_is_enabled(struct tb_port *port)
 {
-	u32 data;
+	u32 data[2];
 
-	if (tb_port_read(port, &data, TB_CFG_PORT, port->cap_adap, 1))
+	if (tb_port_read(port, data, TB_CFG_PORT, port->cap_adap,
+			 ARRAY_SIZE(data)))
 		return false;
 
-	return !!(data & (TB_DP_VIDEO_EN | TB_DP_AUX_EN));
+	return !!(data[0] & (TB_DP_VIDEO_EN | TB_DP_AUX_EN));
 }
 
 /**
@@ -914,19 +915,21 @@ bool tb_dp_port_is_enabled(struct tb_port *port)
  */
 int tb_dp_port_enable(struct tb_port *port, bool enable)
 {
-	u32 data;
+	u32 data[2];
 	int ret;
 
-	ret = tb_port_read(port, &data, TB_CFG_PORT, port->cap_adap, 1);
+	ret = tb_port_read(port, data, TB_CFG_PORT, port->cap_adap,
+			   ARRAY_SIZE(data));
 	if (ret)
 		return ret;
 
 	if (enable)
-		data |= TB_DP_VIDEO_EN | TB_DP_AUX_EN;
+		data[0] |= TB_DP_VIDEO_EN | TB_DP_AUX_EN;
 	else
-		data &= ~(TB_DP_VIDEO_EN | TB_DP_AUX_EN);
+		data[0] &= ~(TB_DP_VIDEO_EN | TB_DP_AUX_EN);
 
-	return tb_port_write(port, &data, TB_CFG_PORT, port->cap_adap, 1);
+	return tb_port_write(port, data, TB_CFG_PORT, port->cap_adap,
+			     ARRAY_SIZE(data));
 }
 
 /* switch utility functions */
Signature · 1 line
2.23.0
Tested-by on v1 1/3
#2 lore raw

Re: [PATCH 1/3] thunderbolt: Read DP IN adapter first two dwords in one go

Quoted 21 lines · Mika Westerberg

When we discover existing DP tunnels the code checks whether DP IN adapter port is enabled by calling tb_dp_port_is_enabled() before it continues the discovery process. On Light Ridge (gen 1) controller reading only the first dword of the DP IN config space causes subsequent access to the same DP IN port path config space to fail or return invalid data as can be seen in the below splat:

   thunderbolt 0000:07:00.0: CFG_ERROR(0:d): Invalid config space or offset
   Call Trace:
    tb_cfg_read+0xb9/0xd0
    __tb_path_deactivate_hop+0x98/0x210
    tb_path_activate+0x228/0x7d0
    tb_tunnel_restart+0x95/0x200
    tb_handle_hotplug+0x30e/0x630
    process_one_work+0x1b4/0x340
    worker_thread+0x44/0x3d0
    kthread+0xeb/0x120
    ? process_one_work+0x340/0x340
    ? kthread_park+0xa0/0xa0
    ret_from_fork+0x1f/0x30
Quoted 7 lines

If both DP In adapter config dwords are read in one go the issue does not reproduce. This is likely firmware bug but we can work it around by always reading the two dwords in one go. There should be no harm for other controllers either so can do it unconditionally.

Quoted 54 lines
---
  drivers/thunderbolt/switch.c | 19 +++++++++++--------
  1 file changed, 11 insertions(+), 8 deletions(-)
drivers/thunderbolt/switch.c +11 -8
index 410bf1bceeee..8e712fbf8233 100644
@@ -896,12 +896,13 @@ int tb_dp_port_set_hops(struct tb_port *port, unsigned int video,
   */
  bool tb_dp_port_is_enabled(struct tb_port *port)
  {
-	u32 data;
+	u32 data[2];
 
-	if (tb_port_read(port, &data, TB_CFG_PORT, port->cap_adap, 1))
+	if (tb_port_read(port, data, TB_CFG_PORT, port->cap_adap,
+			 ARRAY_SIZE(data)))
  		return false;
 
-	return !!(data & (TB_DP_VIDEO_EN | TB_DP_AUX_EN));
+	return !!(data[0] & (TB_DP_VIDEO_EN | TB_DP_AUX_EN));
  }
 
  /**
@@ -914,19 +915,21 @@ bool tb_dp_port_is_enabled(struct tb_port *port)
   */
  int tb_dp_port_enable(struct tb_port *port, bool enable)
  {
-	u32 data;
+	u32 data[2];
  	int ret;
 
-	ret = tb_port_read(port, &data, TB_CFG_PORT, port->cap_adap, 1);
+	ret = tb_port_read(port, data, TB_CFG_PORT, port->cap_adap,
+			   ARRAY_SIZE(data));
  	if (ret)
  		return ret;
 
  	if (enable)
-		data |= TB_DP_VIDEO_EN | TB_DP_AUX_EN;
+		data[0] |= TB_DP_VIDEO_EN | TB_DP_AUX_EN;
  	else
-		data &= ~(TB_DP_VIDEO_EN | TB_DP_AUX_EN);
+		data[0] &= ~(TB_DP_VIDEO_EN | TB_DP_AUX_EN);
 
-	return tb_port_write(port, &data, TB_CFG_PORT, port->cap_adap, 1);
+	return tb_port_write(port, data, TB_CFG_PORT, port->cap_adap,
+			     ARRAY_SIZE(data));
  }
 
  /* switch utility functions */
author Applied
#5 lore raw
In reply to Mika Westerberg
Quoted 4 lines · Mika Westerberg

Mika Westerberg (3):
thunderbolt: Read DP IN adapter first two dwords in one go
thunderbolt: Fix lockdep circular locking depedency warning
thunderbolt: Drop unnecessary read when writing LC command in Ice Lake

Applied to thunderbolt.git/fixes.