Date: Thu, 25 May 2023 15:27:03 +0000
From: Manish Chopra <manishc@...vell.com>
To: Jiri Pirko <jiri@...nulli.us>
CC: "kuba@...nel.org" <kuba@...nel.org>,
        "netdev@...r.kernel.org"
	<netdev@...r.kernel.org>,
        Ariel Elior <aelior@...vell.com>, Alok Prasad
	<palok@...vell.com>,
        Sudarsana Reddy Kalluru <skalluru@...vell.com>,
        David
 Miller <davem@...emloft.net>
Subject: RE: [EXT] Re: [PATCH v5 net] qede: Fix scheduling while atomic

Hi Jiri,

> -----Original Message-----
> From: Jiri Pirko <jiri@...nulli.us>
> Sent: Wednesday, May 24, 2023 5:01 PM
> To: Manish Chopra <manishc@...vell.com>
> Cc: kuba@...nel.org; netdev@...r.kernel.org; Ariel Elior
> <aelior@...vell.com>; Alok Prasad <palok@...vell.com>; Sudarsana Reddy
> Kalluru <skalluru@...vell.com>; David Miller <davem@...emloft.net>
> Subject: [EXT] Re: [PATCH v5 net] qede: Fix scheduling while atomic
> 
> External Email
> 
> ----------------------------------------------------------------------
> Tue, May 23, 2023 at 04:42:35PM CEST, manishc@...vell.com wrote:
> >Bonding module collects the statistics while holding the spinlock,
> >beneath that qede->qed driver statistics flow gets scheduled out due to
> >usleep_range() used in PTT acquire logic which results into below bug
> >and traces -
> >
> >[ 3673.988874] Hardware name: HPE ProLiant DL365 Gen10 Plus/ProLiant
> >DL365 Gen10 Plus, BIOS A42 10/29/2021 [ 3673.988878] Call Trace:
> >[ 3673.988891]  dump_stack_lvl+0x34/0x44 [ 3673.988908]
> >__schedule_bug.cold+0x47/0x53 [ 3673.988918]  __schedule+0x3fb/0x560 [
> >3673.988929]  schedule+0x43/0xb0 [ 3673.988932]
> >schedule_hrtimeout_range_clock+0xbf/0x1b0
> >[ 3673.988937]  ? __hrtimer_init+0xc0/0xc0 [ 3673.988950]
> >usleep_range+0x5e/0x80 [ 3673.988955]  qed_ptt_acquire+0x2b/0xd0 [qed]
> >[ 3673.988981]  _qed_get_vport_stats+0x141/0x240 [qed] [ 3673.989001]
> >qed_get_vport_stats+0x18/0x80 [qed] [ 3673.989016]
> >qede_fill_by_demand_stats+0x37/0x400 [qede] [ 3673.989028]
> >qede_get_stats64+0x19/0xe0 [qede] [ 3673.989034]
> >dev_get_stats+0x5c/0xc0 [ 3673.989045]
> >netstat_show.constprop.0+0x52/0xb0
> >[ 3673.989055]  dev_attr_show+0x19/0x40 [ 3673.989065]
> >sysfs_kf_seq_show+0x9b/0xf0 [ 3673.989076]  seq_read_iter+0x120/0x4b0 [
> >3673.989087]  new_sync_read+0x118/0x1a0 [ 3673.989095]
> >vfs_read+0xf3/0x180 [ 3673.989099]  ksys_read+0x5f/0xe0 [ 3673.989102]
> >do_syscall_64+0x3b/0x90 [ 3673.989109]
> >entry_SYSCALL_64_after_hwframe+0x44/0xae
> 
> You mention "bonding module" at the beginning of this description. Where
> exactly is that shown in the trace?
> 
> I guess that the "spinlock" you talk about is "dev_base_lock", isn't it?

Bonding function somehow were not part of traces, but this is the flow from bonding module
which calls dev_get_stats() under spin_lock_nested(&bond->stats_lock, nest_level) which results to this issue.

> 
> 
> >[ 3673.989115] RIP: 0033:0x7f8467d0b082 [ 3673.989119] Code: c0 e9 b2
> >fe ff ff 50 48 8d 3d ca 05 08 00 e8 35 e7 01 00 0f 1f 44 00 00 f3 0f 1e
> >fa 64 8b 04 25 18 00 00 00 85 c0 75 10 0f 05 <48> 3d 00 f0 ff ff 77 56
> >c3 0f 1f 44 00 00 48 83 ec 28 48 89 54 24 [ 3673.989121] RSP:
> >002b:00007ffffb21fd08 EFLAGS: 00000246 ORIG_RAX: 0000000000000000 [
> >3673.989127] RAX: ffffffffffffffda RBX: 000000000100eca0 RCX:
> >00007f8467d0b082 [ 3673.989128] RDX: 00000000000003ff RSI:
> >00007ffffb21fdc0 RDI: 0000000000000003 [ 3673.989130] RBP:
> 00007f8467b96028 R08: 0000000000000010 R09: 00007ffffb21ec00 [
> 3673.989132] R10: 00007ffffb27b170 R11: 0000000000000246 R12:
> 00000000000000f0 [ 3673.989134] R13: 0000000000000003 R14:
> 00007f8467b92000 R15: 0000000000045a05
> >[ 3673.989139] CPU: 30 PID: 285188 Comm: read_all Kdump: loaded
> Tainted: G        W  OE
> >
> >Fix this by collecting the statistics asynchronously from a periodic
> >delayed work scheduled at default stats coalescing interval and return
> >the recent copy of statisitcs from .ndo_get_stats64(), also add ability
> >to configure/retrieve stats coalescing interval using below commands -
> >
> >ethtool -C ethx stats-block-usecs <val> ethtool -c ethx
> >
> >Fixes: 133fac0eedc3 ("qede: Add basic ethtool support")
> >Cc: Sudarsana Kalluru <skalluru@...vell.com>
> >Cc: David Miller <davem@...emloft.net>
> >Signed-off-by: Manish Chopra <manishc@...vell.com>
> >---
> >v1->v2:
> > - Fixed checkpatch and kdoc warnings.
> >v2->v3:
> > - Moving the changelog after tags.
> >v3->v4:
> > - Changes to collect stats periodically using delayed work
> >   and add ability to configure/retrieve stats coalescing
> >   interval using ethtool
> > - Modified commit description to reflect the changes
> >v4->v5:
> > - Renamed the variables (s/ticks/usecs and s/interval/ticks)
> > - Relaxed the stats usecs coalescing configuration to allow
> >   user to set any range of values and also while getting return
> >   the exact value configured
> > - Usage of usecs_to_jiffies() wherever applicable
> > - Cosmetic change for logs/comments
> >---
> > drivers/net/ethernet/qlogic/qede/qede.h       |  4 +++
> > .../net/ethernet/qlogic/qede/qede_ethtool.c   | 26 ++++++++++++--
> > drivers/net/ethernet/qlogic/qede/qede_main.c  | 35 ++++++++++++++++++-
> > 3 files changed, 62 insertions(+), 3 deletions(-)
> >
> >diff --git a/drivers/net/ethernet/qlogic/qede/qede.h
> >b/drivers/net/ethernet/qlogic/qede/qede.h
> >index f90dcfe9ee68..8a63f99d499c 100644
> >--- a/drivers/net/ethernet/qlogic/qede/qede.h
> >+++ b/drivers/net/ethernet/qlogic/qede/qede.h
> >@@ -271,6 +271,10 @@ struct qede_dev {
> > #define QEDE_ERR_WARN			3
> >
> > 	struct qede_dump_info		dump_info;
> >+	struct delayed_work		periodic_task;
> >+	unsigned long			stats_coal_ticks;
> >+	u32				stats_coal_usecs;
> >+	spinlock_t			stats_lock; /* lock for vport stats
> access */
> > };
> >
> > enum QEDE_STATE {
> >diff --git a/drivers/net/ethernet/qlogic/qede/qede_ethtool.c
> >b/drivers/net/ethernet/qlogic/qede/qede_ethtool.c
> >index 8284c4c1528f..a6498eb7cbd7 100644
> >--- a/drivers/net/ethernet/qlogic/qede/qede_ethtool.c
> >+++ b/drivers/net/ethernet/qlogic/qede/qede_ethtool.c
> >@@ -426,6 +426,8 @@ static void qede_get_ethtool_stats(struct net_device
> *dev,
> > 		}
> > 	}
> >
> >+	spin_lock(&edev->stats_lock);
> >+
> > 	for (i = 0; i < QEDE_NUM_STATS; i++) {
> > 		if (qede_is_irrelevant_stat(edev, i))
> > 			continue;
> >@@ -435,6 +437,8 @@ static void qede_get_ethtool_stats(struct net_device
> *dev,
> > 		buf++;
> > 	}
> >
> >+	spin_unlock(&edev->stats_lock);
> >+
> > 	__qede_unlock(edev);
> > }
> >
> >@@ -817,6 +821,7 @@ static int qede_get_coalesce(struct net_device
> >*dev,
> >
> > 	coal->rx_coalesce_usecs = rx_coal;
> > 	coal->tx_coalesce_usecs = tx_coal;
> >+	coal->stats_block_coalesce_usecs = edev->stats_coal_usecs;
> >
> > 	return rc;
> > }
> >@@ -830,6 +835,21 @@ int qede_set_coalesce(struct net_device *dev,
> struct ethtool_coalesce *coal,
> > 	int i, rc = 0;
> > 	u16 rxc, txc;
> >
> >+	if (edev->stats_coal_usecs != coal->stats_block_coalesce_usecs) {
> >+		bool stats_coal_enabled;
> >+
> >+		stats_coal_enabled = edev->stats_coal_usecs ? true : false;
> >+
> >+		edev->stats_coal_usecs = coal->stats_block_coalesce_usecs;
> >+		edev->stats_coal_ticks =
> >+usecs_to_jiffies(coal->stats_block_coalesce_usecs);
> >+
> >+		if (!stats_coal_enabled)
> >+			schedule_delayed_work(&edev->periodic_task, 0);
> 
> What is the point of schedule here? Don't you want to rather schedule if
> (stats_coal_enabled == true) ??

This was for scheduling of periodic task ONLY if previously it was disabled.
But actually it does not harm to schedule once always whenever user sets a non-zero usecs.

> 
> 
> >+
> >+		DP_INFO(edev, "Configured stats coal ticks=%lu jiffies\n",
> >+			edev->stats_coal_ticks);
> >+	}
> >+
> > 	if (!netif_running(dev)) {
> > 		DP_INFO(edev, "Interface is down\n");
> > 		return -EINVAL;
> >@@ -2236,7 +2256,8 @@ static int qede_get_per_coalesce(struct
> >net_device *dev,  }
> >
> > static const struct ethtool_ops qede_ethtool_ops = {
> >-	.supported_coalesce_params	= ETHTOOL_COALESCE_USECS,
> >+	.supported_coalesce_params	= ETHTOOL_COALESCE_USECS |
> >+
> ETHTOOL_COALESCE_STATS_BLOCK_USECS,
> > 	.get_link_ksettings		= qede_get_link_ksettings,
> > 	.set_link_ksettings		= qede_set_link_ksettings,
> > 	.get_drvinfo			= qede_get_drvinfo,
> >@@ -2287,7 +2308,8 @@ static const struct ethtool_ops qede_ethtool_ops
> >= {  };
> >
> > static const struct ethtool_ops qede_vf_ethtool_ops = {
> >-	.supported_coalesce_params	= ETHTOOL_COALESCE_USECS,
> >+	.supported_coalesce_params	= ETHTOOL_COALESCE_USECS |
> >+
> ETHTOOL_COALESCE_STATS_BLOCK_USECS,
> > 	.get_link_ksettings		= qede_get_link_ksettings,
> > 	.get_drvinfo			= qede_get_drvinfo,
> > 	.get_msglevel			= qede_get_msglevel,
> >diff --git a/drivers/net/ethernet/qlogic/qede/qede_main.c
> >b/drivers/net/ethernet/qlogic/qede/qede_main.c
> >index 06c6a5813606..61cc10968988 100644
> >--- a/drivers/net/ethernet/qlogic/qede/qede_main.c
> >+++ b/drivers/net/ethernet/qlogic/qede/qede_main.c
> >@@ -308,6 +308,8 @@ void qede_fill_by_demand_stats(struct qede_dev
> >*edev)
> >
> > 	edev->ops->get_vport_stats(edev->cdev, &stats);
> >
> >+	spin_lock(&edev->stats_lock);
> >+
> > 	p_common->no_buff_discards = stats.common.no_buff_discards;
> > 	p_common->packet_too_big_discard =
> stats.common.packet_too_big_discard;
> > 	p_common->ttl0_discard = stats.common.ttl0_discard; @@ -405,6
> +407,8
> >@@ void qede_fill_by_demand_stats(struct qede_dev *edev)
> > 		p_ah->tx_1519_to_max_byte_packets =
> > 		    stats.ah.tx_1519_to_max_byte_packets;
> > 	}
> >+
> >+	spin_unlock(&edev->stats_lock);
> > }
> >
> > static void qede_get_stats64(struct net_device *dev, @@ -413,9 +417,10
> >@@ static void qede_get_stats64(struct net_device *dev,
> > 	struct qede_dev *edev = netdev_priv(dev);
> > 	struct qede_stats_common *p_common;
> >
> >-	qede_fill_by_demand_stats(edev);
> > 	p_common = &edev->stats.common;
> >
> >+	spin_lock(&edev->stats_lock);
> >+
> > 	stats->rx_packets = p_common->rx_ucast_pkts + p_common-
> >rx_mcast_pkts +
> > 			    p_common->rx_bcast_pkts;
> > 	stats->tx_packets = p_common->tx_ucast_pkts + p_common-
> >tx_mcast_pkts
> >+ @@ -435,6 +440,8 @@ static void qede_get_stats64(struct net_device
> *dev,
> > 		stats->collisions = edev->stats.bb.tx_total_collisions;
> > 	stats->rx_crc_errors = p_common->rx_crc_errors;
> > 	stats->rx_frame_errors = p_common->rx_align_errors;
> >+
> >+	spin_unlock(&edev->stats_lock);
> > }
> >
> > #ifdef CONFIG_QED_SRIOV
> >@@ -1000,6 +1007,21 @@ static void qede_unlock(struct qede_dev *edev)
> > 	rtnl_unlock();
> > }
> >
> >+static void qede_periodic_task(struct work_struct *work) {
> >+	struct qede_dev *edev = container_of(work, struct qede_dev,
> >+					     periodic_task.work);
> >+
> >+	if (test_bit(QEDE_SP_DISABLE, &edev->sp_flags))
> >+		return;
> >+
> >+	if (edev->stats_coal_usecs) {
> 
> Why don't you cancel the work when you don't want this to happen?

Ok.
  
> 
> 
> >+		qede_fill_by_demand_stats(edev);
> >+		schedule_delayed_work(&edev->periodic_task,
> >+				      edev->stats_coal_ticks);
> >+	}
> >+}
> >+
> > static void qede_sp_task(struct work_struct *work)  {
> > 	struct qede_dev *edev = container_of(work, struct qede_dev, @@
> >-1208,7 +1230,9 @@ static int __qede_probe(struct pci_dev *pdev, u32
> dp_module, u8 dp_level,
> > 		 * from there, although it's unlikely].
> > 		 */
> > 		INIT_DELAYED_WORK(&edev->sp_task, qede_sp_task);
> >+		INIT_DELAYED_WORK(&edev->periodic_task,
> qede_periodic_task);
> > 		mutex_init(&edev->qede_lock);
> >+		spin_lock_init(&edev->stats_lock);
> >
> > 		rc = register_netdev(edev->ndev);
> > 		if (rc) {
> >@@ -1233,6 +1257,11 @@ static int __qede_probe(struct pci_dev *pdev,
> u32 dp_module, u8 dp_level,
> > 	edev->rx_copybreak = QEDE_RX_HDR_SIZE;
> >
> > 	qede_log_probe(edev);
> >+
> >+	edev->stats_coal_usecs = USEC_PER_SEC;
> >+	edev->stats_coal_ticks = usecs_to_jiffies(USEC_PER_SEC);
> >+	schedule_delayed_work(&edev->periodic_task, 0);
> >+
> > 	return 0;
> >
> > err4:
> >@@ -1301,6 +1330,7 @@ static void __qede_remove(struct pci_dev *pdev,
> enum qede_remove_mode mode)
> > 		unregister_netdev(ndev);
> >
> > 		cancel_delayed_work_sync(&edev->sp_task);
> >+		cancel_delayed_work_sync(&edev->periodic_task);
> >
> > 		edev->ops->common->set_power_state(cdev, PCI_D0);
> >
> >@@ -2571,6 +2601,9 @@ static void qede_recovery_handler(struct
> qede_dev
> >*edev)
> >
> > 	DP_NOTICE(edev, "Starting a recovery process\n");
> >
> >+	/* disable periodic stats */
> >+	edev->stats_coal_usecs = 0;
> 
> You disable but never enable again. Why?
> 

It's getting enabled back when recovery flow completes in below part of changes in the patch.

+	edev->stats_coal_usecs = USEC_PER_SEC;
+	edev->stats_coal_ticks = usecs_to_jiffies(USEC_PER_SEC);
+	schedule_delayed_work(&edev->periodic_task, 0);

Although on reset I am resetting values to default load values but maybe I will cleanup here to retain ethtool set values.

> Also, why don't you do:
> cancel_delayed_work_sync(&edev->periodic_task)
> here instead?
> 

Sure.

> 
> >+
> > 	/* No need to acquire first the qede_lock since is done by
> qede_sp_task
> > 	 * before calling this function.
> > 	 */
> >--
> >2.27.0
> >
> >

Powered by blists - more mailing lists