netdev - [PATCH net 3/7] i40e: Fix changing previously set num_queue

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20211118003159.245561-4-anthony.l.nguyen@intel.com>
Date:   Wed, 17 Nov 2021 16:31:55 -0800
From:   Tony Nguyen <anthony.l.nguyen@...el.com>
To:     davem@...emloft.net, kuba@...nel.org
Cc:     Eryk Rybak <eryk.roch.rybak@...el.com>, netdev@...r.kernel.org,
        anthony.l.nguyen@...el.com, sassmann@...hat.com,
        Maciej Fijalkowski <maciej.fijalkowski@...el.com>,
        Przemyslaw Patynowski <przemyslawx.patynowski@...el.com>,
        Tony Brelinski <tony.brelinski@...el.com>
Subject: [PATCH net 3/7] i40e: Fix changing previously set num_queue_pairs for PFs

From: Eryk Rybak <eryk.roch.rybak@...el.com>

Currently, the i40e_vsi_setup_queue_map is basing the count of queues in
TCs on a VSI's alloc_queue_pairs member which is not changed throughout
any user's action (for example via ethtool's set_channels callback).

This implies that vsi->tc_config.tc_info[n].qcount value that is given
to the kernel via netdev_set_tc_queue() that notifies about the count of
queues per particular traffic class is constant even if user has changed
the total count of queues.

This in turn caused the kernel warning after setting the queue count to
the lower value than the initial one:

$ ethtool -l ens801f0
Channel parameters for ens801f0:
Pre-set maximums:
RX:             0
TX:             0
Other:          1
Combined:       64
Current hardware settings:
RX:             0
TX:             0
Other:          1
Combined:       64

$ ethtool -L ens801f0 combined 40

[dmesg]
Number of in use tx queues changed invalidating tc mappings. Priority
traffic classification disabled!

Reason was that vsi->alloc_queue_pairs stayed at 64 value which was used
to set the qcount on TC0 (by default only TC0 exists so all of the
existing queues are assigned to TC0). we update the offset/qcount via
netdev_set_tc_queue() back to the old value but then the
netif_set_real_num_tx_queues() is using the vsi->num_queue_pairs as a
value which got set to 40.

Fix it by using vsi->req_queue_pairs as a queue count that will be
distributed across TCs. Do it only for non-zero values, which implies
that user actually requested the new count of queues.

For VSIs other than main, stay with the vsi->alloc_queue_pairs as we
only allow manipulating the queue count on main VSI.

Fixes: bc6d33c8d93f ("i40e: Fix the number of queues available to be mapped for use")
Co-developed-by: Maciej Fijalkowski <maciej.fijalkowski@...el.com>
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@...el.com>
Co-developed-by: Przemyslaw Patynowski <przemyslawx.patynowski@...el.com>
Signed-off-by: Przemyslaw Patynowski <przemyslawx.patynowski@...el.com>
Signed-off-by: Eryk Rybak <eryk.roch.rybak@...el.com>
Tested-by: Tony Brelinski <tony.brelinski@...el.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@...el.com>
---
 drivers/net/ethernet/intel/i40e/i40e_main.c | 35 ++++++++++++++-------
 1 file changed, 23 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index 6e309d6ce37d..8437cc14bfc6 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -1790,6 +1790,7 @@ static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
 				     bool is_add)
 {
 	struct i40e_pf *pf = vsi->back;
+	u16 num_tc_qps = 0;
 	u16 sections = 0;
 	u8 netdev_tc = 0;
 	u16 numtc = 1;
@@ -1797,13 +1798,29 @@ static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
 	u8 offset;
 	u16 qmap;
 	int i;
-	u16 num_tc_qps = 0;
 
 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
 	offset = 0;
 
+	if (vsi->type == I40E_VSI_MAIN) {
+		/* This code helps add more queue to the VSI if we have
+		 * more cores than RSS can support, the higher cores will
+		 * be served by ATR or other filters. Furthermore, the
+		 * non-zero req_queue_pairs says that user requested a new
+		 * queue count via ethtool's set_channels, so use this
+		 * value for queues distribution across traffic classes
+		 */
+		if (vsi->req_queue_pairs > 0)
+			vsi->num_queue_pairs = vsi->req_queue_pairs;
+		else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
+			vsi->num_queue_pairs = pf->num_lan_msix;
+	}
+
 	/* Number of queues per enabled TC */
-	num_tc_qps = vsi->alloc_queue_pairs;
+	if (vsi->type == I40E_VSI_MAIN)
+		num_tc_qps = vsi->num_queue_pairs;
+	else
+		num_tc_qps = vsi->alloc_queue_pairs;
 	if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
 		/* Find numtc from enabled TC bitmap */
 		for (i = 0, numtc = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
@@ -1881,16 +1898,10 @@ static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
 		}
 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
 	}
-
-	/* Set actual Tx/Rx queue pairs */
-	vsi->num_queue_pairs = offset;
-	if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
-		if (vsi->req_queue_pairs > 0)
-			vsi->num_queue_pairs = vsi->req_queue_pairs;
-		else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
-			vsi->num_queue_pairs = pf->num_lan_msix;
-	}
-
+	/* Do not change previously set num_queue_pairs for PFs */
+	if ((vsi->type == I40E_VSI_MAIN && numtc != 1) ||
+	    vsi->type != I40E_VSI_MAIN)
+		vsi->num_queue_pairs = offset;
 	/* Scheduler section valid can only be set for ADD VSI */
 	if (is_add) {
 		sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
-- 
2.31.1