fix_regression_1

From pasic@linux.ibm.com Fri Dec 12 13:19:27 2025
X-Mozilla-Status: 0001
X-Mozilla-Status2: 00000000
Return-Path: <pasic@linux.ibm.com>
Received: from ltc-imap-p9a.rchland.ibm.com ([unix socket])
	 by ltc-imap-p9a.rchland.ibm.com (Cyrus 3.0.7-27.el8_10 Fedora) with LMTPA;
	 Fri, 12 Dec 2025 07:19:42 -0600
X-Cyrus-Session-Id: ltc-imap-p9a.rchland.ibm.com-2726422-1765545582-2-2378096810068351409
X-Sieve: CMU Sieve 3.0
Received: from localhost (localhost.localdomain [127.0.0.1])
	by ltc-imap-p9a.rchland.ibm.com (Postfix) with ESMTP id 1DB15818A6868;
	Fri, 12 Dec 2025 07:19:42 -0600 (CST)
X-Virus-Scanned: amavis at linux.ibm.com
X-Spam-Flag: NO
X-Spam-Score: 0.001
X-Spam-Level:
X-Spam-Status: No, score=0.001 tagged_above=-9999 required=6.2
 tests=[SPF_HELO_NONE=0.001] autolearn=disabled
Received: from ltc-imap-p9a.rchland.ibm.com ([127.0.0.1])
 by localhost (ltc-imap-p9a.rchland.ibm.com [127.0.0.1]) (amavis, port 10024)
 with LMTP id kaWPjPH15vuw; Fri, 12 Dec 2025 07:19:40 -0600 (CST)
Received: from mx2.linux.ibm.com (ltc-mx2-p9a.rchland.ibm.com [9.5.196.148])
	by imap.linux.ibm.com (Postfix) with ESMTPS id 0056B818877BC;
	Fri, 12 Dec 2025 07:19:40 -0600 (CST)
Received: from smtprelay03.fra02v.mail.ibm.com (unknown [9.218.2.224])
	by mx2.linux.ibm.com (Postfix) with ESMTPS id 80B47207DCE;
	Fri, 12 Dec 2025 07:19:39 -0600 (CST)
Received: from smtpav03.fra02v.mail.ibm.com (smtpav03.fra02v.mail.ibm.com [10.20.54.102])
	by smtprelay03.fra02v.mail.ibm.com (8.14.9/8.14.9/NCO v10.0) with ESMTP id 5BCDJaQ854985054
	(version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-GCM-SHA384 bits=256 verify=OK);
	Fri, 12 Dec 2025 13:19:36 GMT
Received: from smtpav03.fra02v.mail.ibm.com (unknown [127.0.0.1])
	by IMSVA (Postfix) with ESMTP id 488F720043;
	Fri, 12 Dec 2025 13:19:36 +0000 (GMT)
Received: from smtpav03.fra02v.mail.ibm.com (unknown [127.0.0.1])
	by IMSVA (Postfix) with ESMTP id 1FE6B20040;
	Fri, 12 Dec 2025 13:19:36 +0000 (GMT)
Received: from tuxmaker.boeblingen.de.ibm.com (unknown [9.87.85.9])
	by smtpav03.fra02v.mail.ibm.com (Postfix) with ESMTP;
	Fri, 12 Dec 2025 13:19:36 +0000 (GMT)
From: Halil Pasic <pasic@linux.ibm.com>
To: Sidraya Jayagond <sidraya@linux.ibm.com>,
        Mahanta Jambigi <mjambigi@linux.ibm.com>,
        linux390-list@tuxmaker.boeblingen.de.ibm.com
Cc: Halil Pasic <pasic@linux.ibm.com>,
        Alexandra Winter <wintera@linux.ibm.com>, hidayath@linux.ibm.com,
        aswin@linux.ibm.com, wenjia@linux.ibm.com, gbayer@linux.ibm.com,
        nagamani@linux.ibm.com
Subject: [PATCH v2 1/1] smc_rnics: fix regression when PFT not available
Date: Fri, 12 Dec 2025 14:19:27 +0100
Message-ID: <20251212131927.2667798-1-pasic@linux.ibm.com>
X-Mailer: git-send-email 2.51.0
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
X-TM-AS-GCONF: 00

Before commit 0b965ae ("smc_rnics: Add support for Network Express RNIC
in smc_rnics") the device type detection was based on PCI vendor and
device ID. The aforementioned commit switched smc_rnics to doing
the device type detection primarily on PCI function type (PFT), which
is in general a sound thing to do.

In certain vitalized environments however (most notably QEMU+KVM)
PFT is generally not available, and even for passed through devices,
as of now PFT 0 is reported regardless of the actual PFT as seen
by G1. And that leads to not detecting passed thorough rnics as such,
and generally displaying an empty list when smc_rnics is called with
no parameters, even in a presence of rnics to display.

The patch also degraded our ability to handle to the script unknown
vendor id 0x15b3 devices gracefully. This can be relevant if a not
enlightened version of the script is used on new hardware.

Adding a warning about the surprise PFT 0 does not seem to be a
good idea, because there are PCI devices which are readily
available on  s390x and for which PFT 0 is just expected: most
notably virtio-pci, and frankly it isn't all that helpful either.

On the other hand, using the PFT, when available and suitable, is still
considered superior.

Use the old PCI vendor and device ID based logic back as a fall back,
when PFT is not available or not known.

With this the regressions introduced by commit 0b965ae ("smc_rnics: Add
support for Network Express RNIC in smc_rnics") shall be gone.

While at it fix the handling of ports too. The call in the loop that
iterates over network interfaces to set_RoCE_dev_and_port does not make
sense, and 'RoCE Express' needs special handling as a device that has
multiple ports and interfaces per PCI function regardless of the value
of rawIDs. I don't quite understand why grabbing port from the
interface's dev_port attribute is a bad idea. That precedence was
established by 05d1c7e ("smc_rnics: Fix RoCE Express2 port display").

Signed-off-by: Halil Pasic <pasic@linux.ibm.com>
Fixes: commit 0b965ae ("smc_rnics: Add support for Network Express RNIC in smc_rnics")
---
v1->v2:
* do not print not recognized devices (Sid)
* consider the pci device's port attribute with -r (Sid)
* do special port handling for RoCE Express with -r as well
---
 smc_rnics | 37 ++++++++++++++++++++++++++++++++-----
 1 file changed, 32 insertions(+), 5 deletions(-)

diff --git a/smc_rnics b/smc_rnics
index 70b6f69..3621e27 100755
--- a/smc_rnics
+++ b/smc_rnics
@@ -148,6 +148,26 @@ function set_by_firmware_lvl() {
 	set_RoCE_dev_and_port $name;
 }
 
+function set_by_pciid(){
+
+	case "$vend" in
+	"0x1014" ) # IBM
+		case "$id" in
+		"0x04ed") dev_type="ISM";
+			  int="n/a";;
+		esac;;
+	"0x15b3" ) # Mellanox
+		case "$id" in
+		"0x1003" | \
+		"0x1004") dev_type="RoCE_Express";
+			  multiport_f=1;;
+		"0x1016") set_RoCE_dev_and_port "RoCE_Express2";;
+		"0x101e") set_by_firmware_lvl;;
+		*)	  set_RoCE_dev_and_port "Mlx_$id";;
+		esac;;
+	esac
+}
+
 function print_rnics() {
 	# iterate over slots, as powered-off devices won't show elsewhere
 	for fid in `ls -1 /sys/bus/pci/slots`; do
@@ -192,12 +212,15 @@ function print_rnics() {
 		dev_type="${vend#0x}:${id#0x}";
 		pft=`cat pft`;
 		vfn=`cat vfn`;
+		multiport_f=0;
 		if [ $rawIDs -eq 0 ]; then
+			# prefer PFT for disamgibuating devices
 			case "$pft" in
 			"0x05") dev_type="ISM";
 				int="n/a";
 				set_RoCE_pft_and_vfn "$pft" "$vfn";;
 			"0x02") dev_type="RoCE_Express";
+				multipor_f=1;
 				set_RoCE_pft_and_vfn "$pft" "$vfn";;
 			"0x0a") set_by_firmware_lvl;
 				set_RoCE_pft_and_vfn "$pft" "$vfn";;
@@ -205,15 +228,20 @@ function print_rnics() {
 			"0x0f") set_RoCE_dev_and_port "Network_Express";
 				set_RoCE_pft_and_vfn "$pft" "$vfn";;
 			*)
+				# but if that fails go via the pci IDs
+				set_by_pciid
 				# For unknown PCI vendors, determine VF flag based on VFN value.
 				# This ensures consistent handling even for unrecognized vendor devices
-				[ $all -eq 0 ] && continue
+				[ "$dev_type"  = "${vend#0x}:${id#0x}" -a $all -eq 0 ] && continue
 				if (( 16#${vfn#0x} != 0 )); then
 					vfn="y";
 				else
 					vfn="n";
 				fi;;
-		esac
+			esac
+		else
+			set_RoCE_dev_and_port "$dev_type";
+			[ "0x15b3" = "$vend" -a "0x1004" = "$id" ] &&  multiport_f=1;
 		fi
 		pchid="`cat pchid | sed 's/^0x//'`";
 		pnetids="`cat util_string | sed 's/\x0/\x40/g' | iconv -f IBM-1047 -t ASCII`";
@@ -225,11 +253,10 @@ function print_rnics() {
 				print_rnic;
 				continue;
 			fi
-			# one device can have multiple interfaces (one per port)
 			for int in $interfaces; do
-				set_RoCE_dev_and_port "$dev_type";
+				# one device can have multiple interfaces (one per port)
 				cd /sys/bus/pci/devices/$addr/net/$int;
-				if [ "$((pft))" -eq 2 ] && [ -e dev_port ]; then
+				if [ $multiport_f -eq 1 ] && [ -e dev_port ]; then
 					port=`cat dev_port`;
 				fi
 				print_rnic;

base-commit: fdd121a660497db6916cb50b4d3a382c2e02b606
-- 
2.51.0