Merge remote-tracking branch 'op/main'

package/base-files/files/etc/uci-defaults/12_network-generate-ula

@@ -1,11 +1,7 @@
 [ "$(uci -q get network.globals.ula_prefix)" != "auto" ] && exit 0
 
-r1=$(dd if=/dev/urandom bs=1 count=1 |hexdump -e '1/1 "%02x"')
-r2=$(dd if=/dev/urandom bs=2 count=1 |hexdump -e '2/1 "%02x"')
-r3=$(dd if=/dev/urandom bs=2 count=1 |hexdump -e '2/1 "%02x"')
-
 uci -q batch <<-EOF >/dev/null
-	set network.globals.ula_prefix=fd$r1:$r2:$r3::/48
+	set network.globals.ula_prefix="$(hexdump -vn 5 -e '"fd" 1/1 "%02x:" 2/2 "%x:"' /dev/urandom):/48"
 	commit network
 EOF
 

package/kernel/qca-nss-dp/Makefile

@@ -1,7 +1,7 @@
 include $(TOPDIR)/rules.mk
 
 PKG_NAME:=qca-nss-dp
-PKG_RELEASE:=2
+PKG_RELEASE:=3
 
 PKG_SOURCE_URL:=https://git.codelinaro.org/clo/qsdk/oss/lklm/nss-dp.git
 PKG_SOURCE_PROTO:=git

package/kernel/qca-nss-dp/patches/0011-01-edma_v1-rework-hw_reset-logic-to-permit-rmmod-and-in.patch

@@ -0,0 +1,43 @@
+From c318c90b824c59539bf2e33618e381293398616c Mon Sep 17 00:00:00 2001
+From: Christian Marangi <ansuelsmth@gmail.com>
+Date: Tue, 16 Apr 2024 15:02:49 +0200
+Subject: [PATCH 1/6] edma_v1: rework hw_reset logic to permit rmmod and insmod
+
+Rework hw_reset logic for edma v1 to permit rmmod and insmod by using
+get_exclusive_released variant (assuming the reset control was released)
+and manually acquire and release it.
+
+This permits rmmod and insmod without triggering warning or receiving
+-EBUSY errors.
+
+Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
+---
+ hal/dp_ops/edma_dp/edma_v1/edma_cfg.c | 6 +++++-
+ 1 file changed, 5 insertions(+), 1 deletion(-)
+
+--- a/hal/dp_ops/edma_dp/edma_v1/edma_cfg.c
++++ b/hal/dp_ops/edma_dp/edma_v1/edma_cfg.c
+@@ -719,18 +719,22 @@ int edma_hw_reset(struct edma_hw *ehw)
+ 	struct reset_control *rst;
+ 	struct platform_device *pdev = ehw->pdev;
+
+-	rst = devm_reset_control_get(&pdev->dev, EDMA_HW_RESET_ID);
++	rst = devm_reset_control_get_exclusive_released(&pdev->dev, EDMA_HW_RESET_ID);
+ 	if (IS_ERR(rst)) {
+ 		pr_warn("DTS Node: %s does not exist\n", EDMA_HW_RESET_ID);
+ 		return -EINVAL;
+ 	}
+
++	reset_control_acquire(rst);
++
+ 	reset_control_assert(rst);
+ 	udelay(100);
+
+ 	reset_control_deassert(rst);
+ 	udelay(100);
+
++	reset_control_release(rst);
++
+ 	pr_info("EDMA HW Reset completed succesfully\n");
+
+ 	return 0;

package/kernel/qca-nss-dp/patches/0011-02-nss_dp_switchdev-correctly-unregister-notifier-on-dp.patch

@@ -0,0 +1,59 @@
+From 079bfe441b274a8c06474be82e4ccc88599a5e0e Mon Sep 17 00:00:00 2001
+From: Christian Marangi <ansuelsmth@gmail.com>
+Date: Tue, 16 Apr 2024 16:08:46 +0200
+Subject: [PATCH 2/6] nss_dp_switchdev: correctly unregister notifier on
+ dp_remove
+
+Correctly unregister notifier on dp_remove to fix kernel panic on system
+reboot.
+
+Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
+---
+ include/nss_dp_dev.h |  1 +
+ nss_dp_main.c        |  4 ++++
+ nss_dp_switchdev.c   | 13 +++++++++++++
+ 3 files changed, 18 insertions(+)
+
+--- a/include/nss_dp_dev.h
++++ b/include/nss_dp_dev.h
+@@ -312,6 +312,7 @@ void nss_dp_set_ethtool_ops(struct net_d
+  */
+ #ifdef CONFIG_NET_SWITCHDEV
+ void nss_dp_switchdev_setup(struct net_device *dev);
++void nss_dp_switchdev_remove(struct net_device *dev);
+ bool nss_dp_is_phy_dev(struct net_device *dev);
+ #endif
+
+--- a/nss_dp_main.c
++++ b/nss_dp_main.c
+@@ -913,6 +913,10 @@ static int nss_dp_remove(struct platform
+ 		if (!dp_priv)
+ 			continue;
+
++		#ifdef CONFIG_NET_SWITCHDEV
++			nss_dp_switchdev_remove(dp_priv->netdev);
++		#endif
++
+ 		dp_ops = dp_priv->data_plane_ops;
+ 		hal_ops = dp_priv->gmac_hal_ops;
+
+--- a/nss_dp_switchdev.c
++++ b/nss_dp_switchdev.c
+@@ -635,4 +635,17 @@ void nss_dp_switchdev_setup(struct net_d
+
+ 	switch_init_done = true;
+ }
++
++void nss_dp_switchdev_remove(struct net_device *dev)
++{
++	if (!switch_init_done)
++		return;
++
++	if (nss_dp_sw_ev_nb)
++		unregister_switchdev_notifier(nss_dp_sw_ev_nb);
++
++	unregister_switchdev_blocking_notifier(&nss_dp_switchdev_notifier);
++
++	switch_init_done = false;
++}
+ #endif

package/kernel/qca-nss-dp/patches/0011-03-nss_dp_main-swap-dp_exit-function-call.patch

@@ -0,0 +1,35 @@
+From ab7b1a361d51157118e1a61ce6530a59bcef4b61 Mon Sep 17 00:00:00 2001
+From: Christian Marangi <ansuelsmth@gmail.com>
+Date: Tue, 16 Apr 2024 16:10:09 +0200
+Subject: [PATCH 3/6] nss_dp_main: swap dp_exit function call
+
+First unregister nss_dp platform devices then cleanup the HAL.
+
+This is to fix kernel panic by cleaning data that needs to be used by
+platform driver unregister functions.
+
+Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
+---
+ nss_dp_main.c | 4 ++--
+ 1 file changed, 2 insertions(+), 2 deletions(-)
+
+--- a/nss_dp_main.c
++++ b/nss_dp_main.c
+@@ -1082,6 +1082,8 @@ int __init nss_dp_init(void)
+  */
+ void __exit nss_dp_exit(void)
+ {
++	platform_driver_unregister(&nss_dp_drv);
++
+ 	/*
+ 	 * TODO Move this to soc_ops
+ 	 */
+@@ -1089,8 +1091,6 @@ void __exit nss_dp_exit(void)
+ 		nss_dp_hal_cleanup();
+ 		dp_global_ctx.common_init_done = false;
+ 	}
+-
+-	platform_driver_unregister(&nss_dp_drv);
+ }
+
+ module_init(nss_dp_init);

package/kernel/qca-nss-dp/patches/0011-04-nss_dp_main-call-unregister_netdev-first-in-dp_remov.patch

@@ -0,0 +1,35 @@
+From 33dd3aa6d0f9cd240d63f53a49157ae44ebccf87 Mon Sep 17 00:00:00 2001
+From: Christian Marangi <ansuelsmth@gmail.com>
+Date: Tue, 16 Apr 2024 16:12:11 +0200
+Subject: [PATCH 4/6] nss_dp_main: call unregister_netdev first in dp_remove
+ and carrifer_off
+
+In dp_remove move unregister_netdev up before calling exit and deinit
+and first call netif_carrier_off to stop any traffic from happening and
+prevent kernel panics for napi in the middle of transfer.
+
+Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
+---
+ nss_dp_main.c | 4 +++-
+ 1 file changed, 3 insertions(+), 1 deletion(-)
+
+--- a/nss_dp_main.c
++++ b/nss_dp_main.c
+@@ -920,6 +920,9 @@ static int nss_dp_remove(struct platform
+ 		dp_ops = dp_priv->data_plane_ops;
+ 		hal_ops = dp_priv->gmac_hal_ops;
+
++		netif_carrier_off(dp_priv->netdev);
++		unregister_netdev(dp_priv->netdev);
++
+ 		if (dp_priv->phydev)
+ 			phy_disconnect(dp_priv->phydev);
+
+@@ -931,7 +934,6 @@ static int nss_dp_remove(struct platform
+ #endif
+ 		hal_ops->exit(dp_priv->gmac_hal_ctx);
+ 		dp_ops->deinit(dp_priv->dpc);
+-		unregister_netdev(dp_priv->netdev);
+ 		free_netdev(dp_priv->netdev);
+ 		dp_global_ctx.nss_dp[i] = NULL;
+ 	}

package/kernel/qca-nss-dp/patches/0011-05-nss_dp_main-use-phy_detach-instead-of-disconnect-in-.patch

@@ -0,0 +1,26 @@
+From 655b07b701271bc00952fe64aeb14f993a48a50e Mon Sep 17 00:00:00 2001
+From: Christian Marangi <ansuelsmth@gmail.com>
+Date: Tue, 16 Apr 2024 16:17:36 +0200
+Subject: [PATCH 5/6] nss_dp_main: use phy_detach instead of disconnect in
+ dp_remove
+
+Use phy_detach instead of disconnect in dp_remove. On Module remove, phy
+are already disconnected but they need to be detached to be correctly
+reattached later with an insmod.
+
+Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
+---
+ nss_dp_main.c | 2 +-
+ 1 file changed, 1 insertion(+), 1 deletion(-)
+
+--- a/nss_dp_main.c
++++ b/nss_dp_main.c
+@@ -924,7 +924,7 @@ static int nss_dp_remove(struct platform
+ 		unregister_netdev(dp_priv->netdev);
+
+ 		if (dp_priv->phydev)
+-			phy_disconnect(dp_priv->phydev);
++			phy_detach(dp_priv->phydev);
+
+ #if defined(NSS_DP_PPE_SUPPORT)
+ 		/*

package/kernel/qca-nss-dp/patches/0011-06-edma_v1-skip-edma_disable_port-in-edma_cleanup-subse.patch

@@ -0,0 +1,37 @@
+From c7c59c6097d94dbab8fc68dae798017bdbc5b3b9 Mon Sep 17 00:00:00 2001
+From: Christian Marangi <ansuelsmth@gmail.com>
+Date: Tue, 16 Apr 2024 16:22:32 +0200
+Subject: [PATCH 6/6] edma_v1: skip edma_disable_port in edma_cleanup
+ subsequent run
+
+Skip edma_disable_port in edma_cleanup subsequent run as it will cause
+the kernel panic as the regs are already freed by previous run of
+edma_cleanup. It's use it's not clear but the call is already done in
+the first run of edma_cleanup. Maybe an oversight never dropped?
+
+Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
+---
+ hal/dp_ops/edma_dp/edma_v1/edma_data_plane.c | 12 +++++++++---
+ 1 file changed, 9 insertions(+), 3 deletions(-)
+
+--- a/hal/dp_ops/edma_dp/edma_v1/edma_data_plane.c
++++ b/hal/dp_ops/edma_dp/edma_v1/edma_data_plane.c
+@@ -326,9 +326,15 @@ void edma_cleanup(bool is_dp_override)
+ 		 * Disable EDMA only at module exit time, since NSS firmware
+ 		 * depends on this setting.
+ 		 */
+-		if (!is_dp_override) {
+-			edma_disable_port();
+-		}
++		/* This call will make the kernel panic as reg used by
++		 * edma_disable_port are already freed by previous call of
++		 * edma_cleanup. Logic is not clear of WHY this is called.
++		 * Keep this here for reference if someone EVER wants
++		 * to investigate.
++		 */
++		// if (!is_dp_override) {
++		// 	edma_disable_port();
++		// }
+ 		return;
+ 	}
+

package/libs/ustream-ssl/Makefile

@@ -5,9 +5,9 @@ PKG_RELEASE:=1
 
 PKG_SOURCE_PROTO:=git
 PKG_SOURCE_URL=$(PROJECT_GIT)/project/ustream-ssl.git
-PKG_SOURCE_DATE:=2024-04-07
-PKG_SOURCE_VERSION:=d61493a4420417cbf9931ffee8c862faf04f2967
-PKG_MIRROR_HASH:=a420d775ad4928836e33e0d9423486fe4904555dbbfff283cd96370a49cf9659
+PKG_SOURCE_DATE:=2024-04-18
+PKG_SOURCE_VERSION:=af7c3532ad499bc97f568087a57ac0f4fe6f56e1
+PKG_MIRROR_HASH:=a1aaa6d9245bb7007c070312587844fe4054fd95af3ba3fffd942ba852ff2c8f
 CMAKE_INSTALL:=1
 
 PKG_LICENSE:=ISC

target/linux/qoriq/Makefile

@@ -12,6 +12,7 @@ FEATURES:=boot-part ext4 fpu legacy-sdcard powerpc64 ramdisk rootfs-part rtc sou
 SUBTARGETS:=generic
 
 KERNEL_PATCHVER:=5.15
+KERNEL_TESTING_PATCHVER:=6.1
 
 KERNELNAME:=zImage
 

target/linux/qoriq/config-5.15

@@ -113,7 +113,6 @@ CONFIG_FSL_CORENET_CF=y
 CONFIG_FSL_CORENET_RCPM=y
 CONFIG_FSL_DMA=y
 CONFIG_FSL_DPAA=y
-# CONFIG_FSL_DPAA2_SWITCH is not set
 # CONFIG_FSL_DPAA_CHECKING is not set
 CONFIG_FSL_DPAA_ETH=y
 CONFIG_FSL_EMB_PERFMON=y
@@ -132,6 +131,7 @@ CONFIG_FS_IOMAP=y
 CONFIG_FS_MBCACHE=y
 CONFIG_FS_POSIX_ACL=y
 CONFIG_FTL=y
+CONFIG_FUNCTION_ALIGNMENT=0
 CONFIG_FUNCTION_ERROR_INJECTION=y
 CONFIG_FWNODE_MDIO=y
 CONFIG_FW_LOADER_PAGED_BUF=y