]> git.ipfire.org Git - thirdparty/kernel/stable-queue.git/commitdiff
Fixes for 4.4
authorSasha Levin <sashal@kernel.org>
Fri, 5 Jun 2020 01:12:22 +0000 (21:12 -0400)
committerSasha Levin <sashal@kernel.org>
Fri, 5 Jun 2020 01:12:22 +0000 (21:12 -0400)
Signed-off-by: Sasha Levin <sashal@kernel.org>
queue-4.4/arc-fix-iccm-dccm-runtime-size-checks.patch [new file with mode: 0644]
queue-4.4/net-bmac-fix-read-of-mac-address-from-rom.patch [new file with mode: 0644]
queue-4.4/net-ethernet-freescale-rework-quiesce-activate-for-u.patch [new file with mode: 0644]
queue-4.4/net-ethernet-stmmac-enable-interface-clocks-on-probe.patch [new file with mode: 0644]
queue-4.4/pppoe-only-process-padt-targeted-at-local-interfaces.patch [new file with mode: 0644]
queue-4.4/series
queue-4.4/x86-mmiotrace-use-cpumask_available-for-cpumask_var_.patch [new file with mode: 0644]

diff --git a/queue-4.4/arc-fix-iccm-dccm-runtime-size-checks.patch b/queue-4.4/arc-fix-iccm-dccm-runtime-size-checks.patch
new file mode 100644 (file)
index 0000000..ac673e5
--- /dev/null
@@ -0,0 +1,54 @@
+From 68fe24022c9753da6fac2a60c7f9de1bcb847565 Mon Sep 17 00:00:00 2001
+From: Sasha Levin <sashal@kernel.org>
+Date: Thu, 2 Apr 2020 20:54:28 +0300
+Subject: ARC: Fix ICCM & DCCM runtime size checks
+
+From: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
+
+[ Upstream commit 43900edf67d7ef3ac8909854d75b8a1fba2d570c ]
+
+As of today the ICCM and DCCM size checks are incorrectly using
+mismatched units (KiB checked against bytes). The CONFIG_ARC_DCCM_SZ
+and CONFIG_ARC_ICCM_SZ are in KiB, but the size calculated in
+runtime and stored in cpu->dccm.sz and cpu->iccm.sz is in bytes.
+
+Fix that.
+
+Reported-by: Paul Greco <pmgreco@us.ibm.com>
+Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
+Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
+Signed-off-by: Sasha Levin <sashal@kernel.org>
+---
+ arch/arc/kernel/setup.c | 5 +++--
+ 1 file changed, 3 insertions(+), 2 deletions(-)
+
+diff --git a/arch/arc/kernel/setup.c b/arch/arc/kernel/setup.c
+index 3013f3f82b95..66e4dc8bce1d 100644
+--- a/arch/arc/kernel/setup.c
++++ b/arch/arc/kernel/setup.c
+@@ -12,6 +12,7 @@
+ #include <linux/root_dev.h>
+ #include <linux/console.h>
+ #include <linux/module.h>
++#include <linux/sizes.h>
+ #include <linux/cpu.h>
+ #include <linux/clk-provider.h>
+ #include <linux/of_fdt.h>
+@@ -308,12 +309,12 @@ static void arc_chk_core_config(void)
+       if ((unsigned int)__arc_dccm_base != cpu->dccm.base_addr)
+               panic("Linux built with incorrect DCCM Base address\n");
+-      if (CONFIG_ARC_DCCM_SZ != cpu->dccm.sz)
++      if (CONFIG_ARC_DCCM_SZ * SZ_1K != cpu->dccm.sz)
+               panic("Linux built with incorrect DCCM Size\n");
+ #endif
+ #ifdef CONFIG_ARC_HAS_ICCM
+-      if (CONFIG_ARC_ICCM_SZ != cpu->iccm.sz)
++      if (CONFIG_ARC_ICCM_SZ * SZ_1K != cpu->iccm.sz)
+               panic("Linux built with incorrect ICCM Size\n");
+ #endif
+-- 
+2.25.1
+
diff --git a/queue-4.4/net-bmac-fix-read-of-mac-address-from-rom.patch b/queue-4.4/net-bmac-fix-read-of-mac-address-from-rom.patch
new file mode 100644 (file)
index 0000000..d4664af
--- /dev/null
@@ -0,0 +1,43 @@
+From 3f3389e74ef2739364ddfe31cf13ca43d3a0a94d Mon Sep 17 00:00:00 2001
+From: Sasha Levin <sashal@kernel.org>
+Date: Tue, 19 May 2020 09:05:58 +0800
+Subject: net: bmac: Fix read of MAC address from ROM
+
+From: Jeremy Kerr <jk@ozlabs.org>
+
+[ Upstream commit ef01cee2ee1b369c57a936166483d40942bcc3e3 ]
+
+In bmac_get_station_address, We're reading two bytes at a time from ROM,
+but we do that six times, resulting in 12 bytes of read & writes. This
+means we will write off the end of the six-byte destination buffer.
+
+This change fixes the for-loop to only read/write six bytes.
+
+Based on a proposed fix from Finn Thain <fthain@telegraphics.com.au>.
+
+Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
+Reported-by: Stan Johnson <userm57@yahoo.com>
+Tested-by: Stan Johnson <userm57@yahoo.com>
+Reported-by: Finn Thain <fthain@telegraphics.com.au>
+Signed-off-by: David S. Miller <davem@davemloft.net>
+Signed-off-by: Sasha Levin <sashal@kernel.org>
+---
+ drivers/net/ethernet/apple/bmac.c | 2 +-
+ 1 file changed, 1 insertion(+), 1 deletion(-)
+
+diff --git a/drivers/net/ethernet/apple/bmac.c b/drivers/net/ethernet/apple/bmac.c
+index a65d7a60f116..ffa7e7e6d18d 100644
+--- a/drivers/net/ethernet/apple/bmac.c
++++ b/drivers/net/ethernet/apple/bmac.c
+@@ -1187,7 +1187,7 @@ bmac_get_station_address(struct net_device *dev, unsigned char *ea)
+       int i;
+       unsigned short data;
+-      for (i = 0; i < 6; i++)
++      for (i = 0; i < 3; i++)
+               {
+                       reset_and_select_srom(dev);
+                       data = read_srom(dev, i + EnetAddressOffset/2, SROMAddressBits);
+-- 
+2.25.1
+
diff --git a/queue-4.4/net-ethernet-freescale-rework-quiesce-activate-for-u.patch b/queue-4.4/net-ethernet-freescale-rework-quiesce-activate-for-u.patch
new file mode 100644 (file)
index 0000000..9437ed4
--- /dev/null
@@ -0,0 +1,78 @@
+From edb83f589fe52665af2526693084afca7c17f2d4 Mon Sep 17 00:00:00 2001
+From: Sasha Levin <sashal@kernel.org>
+Date: Wed, 20 May 2020 17:53:50 +0200
+Subject: net/ethernet/freescale: rework quiesce/activate for ucc_geth
+
+From: Valentin Longchamp <valentin@longchamp.me>
+
+[ Upstream commit 79dde73cf9bcf1dd317a2667f78b758e9fe139ed ]
+
+ugeth_quiesce/activate are used to halt the controller when there is a
+link change that requires to reconfigure the mac.
+
+The previous implementation called netif_device_detach(). This however
+causes the initial activation of the netdevice to fail precisely because
+it's detached. For details, see [1].
+
+A possible workaround was the revert of commit
+net: linkwatch: add check for netdevice being present to linkwatch_do_dev
+However, the check introduced in the above commit is correct and shall be
+kept.
+
+The netif_device_detach() is thus replaced with
+netif_tx_stop_all_queues() that prevents any tranmission. This allows to
+perform mac config change required by the link change, without detaching
+the corresponding netdevice and thus not preventing its initial
+activation.
+
+[1] https://lists.openwall.net/netdev/2020/01/08/201
+
+Signed-off-by: Valentin Longchamp <valentin@longchamp.me>
+Acked-by: Matteo Ghidoni <matteo.ghidoni@ch.abb.com>
+Signed-off-by: David S. Miller <davem@davemloft.net>
+Signed-off-by: Sasha Levin <sashal@kernel.org>
+---
+ drivers/net/ethernet/freescale/ucc_geth.c | 13 +++++++------
+ 1 file changed, 7 insertions(+), 6 deletions(-)
+
+diff --git a/drivers/net/ethernet/freescale/ucc_geth.c b/drivers/net/ethernet/freescale/ucc_geth.c
+index 55ac00055977..96a1f62cc148 100644
+--- a/drivers/net/ethernet/freescale/ucc_geth.c
++++ b/drivers/net/ethernet/freescale/ucc_geth.c
+@@ -45,6 +45,7 @@
+ #include <asm/ucc.h>
+ #include <asm/ucc_fast.h>
+ #include <asm/machdep.h>
++#include <net/sch_generic.h>
+ #include "ucc_geth.h"
+@@ -1551,11 +1552,8 @@ static int ugeth_disable(struct ucc_geth_private *ugeth, enum comm_dir mode)
+ static void ugeth_quiesce(struct ucc_geth_private *ugeth)
+ {
+-      /* Prevent any further xmits, plus detach the device. */
+-      netif_device_detach(ugeth->ndev);
+-
+-      /* Wait for any current xmits to finish. */
+-      netif_tx_disable(ugeth->ndev);
++      /* Prevent any further xmits */
++      netif_tx_stop_all_queues(ugeth->ndev);
+       /* Disable the interrupt to avoid NAPI rescheduling. */
+       disable_irq(ugeth->ug_info->uf_info.irq);
+@@ -1568,7 +1566,10 @@ static void ugeth_activate(struct ucc_geth_private *ugeth)
+ {
+       napi_enable(&ugeth->napi);
+       enable_irq(ugeth->ug_info->uf_info.irq);
+-      netif_device_attach(ugeth->ndev);
++
++      /* allow to xmit again  */
++      netif_tx_wake_all_queues(ugeth->ndev);
++      __netdev_watchdog_up(ugeth->ndev);
+ }
+ /* Called every time the controller might need to be made
+-- 
+2.25.1
+
diff --git a/queue-4.4/net-ethernet-stmmac-enable-interface-clocks-on-probe.patch b/queue-4.4/net-ethernet-stmmac-enable-interface-clocks-on-probe.patch
new file mode 100644 (file)
index 0000000..d65ec39
--- /dev/null
@@ -0,0 +1,62 @@
+From bd2d699906cdc1c48308d3e4331bc70ba152f88f Mon Sep 17 00:00:00 2001
+From: Sasha Levin <sashal@kernel.org>
+Date: Thu, 21 May 2020 12:49:34 +0100
+Subject: net: ethernet: stmmac: Enable interface clocks on probe for IPQ806x
+
+From: Jonathan McDowell <noodles@earth.li>
+
+[ Upstream commit a96ac8a0045e3cbe3e5af6d1b3c78c6c2065dec5 ]
+
+The ipq806x_gmac_probe() function enables the PTP clock but not the
+appropriate interface clocks. This means that if the bootloader hasn't
+done so attempting to bring up the interface will fail with an error
+like:
+
+[   59.028131] ipq806x-gmac-dwmac 37600000.ethernet: Failed to reset the dma
+[   59.028196] ipq806x-gmac-dwmac 37600000.ethernet eth1: stmmac_hw_setup: DMA engine initialization failed
+[   59.034056] ipq806x-gmac-dwmac 37600000.ethernet eth1: stmmac_open: Hw setup failed
+
+This patch, a slightly cleaned up version of one posted by Sergey
+Sergeev in:
+
+https://forum.openwrt.org/t/support-for-mikrotik-rb3011uias-rm/4064/257
+
+correctly enables the clock; we have already configured the source just
+before this.
+
+Tested on a MikroTik RB3011.
+
+Signed-off-by: Jonathan McDowell <noodles@earth.li>
+Signed-off-by: David S. Miller <davem@davemloft.net>
+Signed-off-by: Sasha Levin <sashal@kernel.org>
+---
+ drivers/net/ethernet/stmicro/stmmac/dwmac-ipq806x.c | 13 +++++++++++++
+ 1 file changed, 13 insertions(+)
+
+diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-ipq806x.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-ipq806x.c
+index 1fc356c17750..15aabffd21a8 100644
+--- a/drivers/net/ethernet/stmicro/stmmac/dwmac-ipq806x.c
++++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-ipq806x.c
+@@ -328,6 +328,19 @@ static int ipq806x_gmac_probe(struct platform_device *pdev)
+       /* Enable PTP clock */
+       regmap_read(gmac->nss_common, NSS_COMMON_CLK_GATE, &val);
+       val |= NSS_COMMON_CLK_GATE_PTP_EN(gmac->id);
++      switch (gmac->phy_mode) {
++      case PHY_INTERFACE_MODE_RGMII:
++              val |= NSS_COMMON_CLK_GATE_RGMII_RX_EN(gmac->id) |
++                      NSS_COMMON_CLK_GATE_RGMII_TX_EN(gmac->id);
++              break;
++      case PHY_INTERFACE_MODE_SGMII:
++              val |= NSS_COMMON_CLK_GATE_GMII_RX_EN(gmac->id) |
++                              NSS_COMMON_CLK_GATE_GMII_TX_EN(gmac->id);
++              break;
++      default:
++              /* We don't get here; the switch above will have errored out */
++              unreachable();
++      }
+       regmap_write(gmac->nss_common, NSS_COMMON_CLK_GATE, val);
+       if (gmac->phy_mode == PHY_INTERFACE_MODE_SGMII) {
+-- 
+2.25.1
+
diff --git a/queue-4.4/pppoe-only-process-padt-targeted-at-local-interfaces.patch b/queue-4.4/pppoe-only-process-padt-targeted-at-local-interfaces.patch
new file mode 100644 (file)
index 0000000..5157436
--- /dev/null
@@ -0,0 +1,40 @@
+From 53f591d79f719c9a66ea9a97a95d136715fbcb57 Mon Sep 17 00:00:00 2001
+From: Sasha Levin <sashal@kernel.org>
+Date: Thu, 14 May 2020 12:15:39 +0200
+Subject: pppoe: only process PADT targeted at local interfaces
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+From: Guillaume Nault <gnault@redhat.com>
+
+We don't want to disconnect a session because of a stray PADT arriving
+while the interface is in promiscuous mode.
+Furthermore, multicast and broadcast packets make no sense here, so
+only PACKET_HOST is accepted.
+
+Reported-by: David Balažic <xerces9@gmail.com>
+Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
+Signed-off-by: Guillaume Nault <gnault@redhat.com>
+Signed-off-by: David S. Miller <davem@davemloft.net>
+---
+ drivers/net/ppp/pppoe.c | 3 +++
+ 1 file changed, 3 insertions(+)
+
+diff --git a/drivers/net/ppp/pppoe.c b/drivers/net/ppp/pppoe.c
+index 39976892b312..13f8464741a1 100644
+--- a/drivers/net/ppp/pppoe.c
++++ b/drivers/net/ppp/pppoe.c
+@@ -494,6 +494,9 @@ static int pppoe_disc_rcv(struct sk_buff *skb, struct net_device *dev,
+       if (!skb)
+               goto out;
++      if (skb->pkt_type != PACKET_HOST)
++              goto abort;
++
+       if (!pskb_may_pull(skb, sizeof(struct pppoe_hdr)))
+               goto abort;
+-- 
+2.25.1
+
index 56e7445f3a1af7340dff7dbcc8c0fc3f935c5448..0f97b7c2d8b410069a3ce1402a268ea3f10a966f 100644 (file)
@@ -6,3 +6,9 @@ ib-mlx4-fix-an-error-handling-path-in-mlx4_ib_rereg_user_mr.patch
 alsa-hda-no-loopback-on-alc299-codec.patch
 spi-dw-use-smp_mb-to-avoid-sending-spi-data-error.patch
 s390-ftrace-save-traced-function-caller.patch
+arc-fix-iccm-dccm-runtime-size-checks.patch
+x86-mmiotrace-use-cpumask_available-for-cpumask_var_.patch
+net-bmac-fix-read-of-mac-address-from-rom.patch
+net-ethernet-freescale-rework-quiesce-activate-for-u.patch
+net-ethernet-stmmac-enable-interface-clocks-on-probe.patch
+pppoe-only-process-padt-targeted-at-local-interfaces.patch
diff --git a/queue-4.4/x86-mmiotrace-use-cpumask_available-for-cpumask_var_.patch b/queue-4.4/x86-mmiotrace-use-cpumask_available-for-cpumask_var_.patch
new file mode 100644 (file)
index 0000000..1ac11b1
--- /dev/null
@@ -0,0 +1,67 @@
+From 144bd9f8dd9141f350e5699c183f71e4cc7eb0f6 Mon Sep 17 00:00:00 2001
+From: Sasha Levin <sashal@kernel.org>
+Date: Wed, 8 Apr 2020 13:53:23 -0700
+Subject: x86/mmiotrace: Use cpumask_available() for cpumask_var_t variables
+
+From: Nathan Chancellor <natechancellor@gmail.com>
+
+[ Upstream commit d7110a26e5905ec2fe3fc88bc6a538901accb72b ]
+
+When building with Clang + -Wtautological-compare and
+CONFIG_CPUMASK_OFFSTACK unset:
+
+  arch/x86/mm/mmio-mod.c:375:6: warning: comparison of array 'downed_cpus'
+  equal to a null pointer is always false [-Wtautological-pointer-compare]
+          if (downed_cpus == NULL &&
+              ^~~~~~~~~~~    ~~~~
+  arch/x86/mm/mmio-mod.c:405:6: warning: comparison of array 'downed_cpus'
+  equal to a null pointer is always false [-Wtautological-pointer-compare]
+          if (downed_cpus == NULL || cpumask_weight(downed_cpus) == 0)
+              ^~~~~~~~~~~    ~~~~
+  2 warnings generated.
+
+Commit
+
+  f7e30f01a9e2 ("cpumask: Add helper cpumask_available()")
+
+added cpumask_available() to fix warnings of this nature. Use that here
+so that clang does not warn regardless of CONFIG_CPUMASK_OFFSTACK's
+value.
+
+Reported-by: Sedat Dilek <sedat.dilek@gmail.com>
+Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
+Signed-off-by: Borislav Petkov <bp@suse.de>
+Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
+Acked-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
+Link: https://github.com/ClangBuiltLinux/linux/issues/982
+Link: https://lkml.kernel.org/r/20200408205323.44490-1-natechancellor@gmail.com
+Signed-off-by: Sasha Levin <sashal@kernel.org>
+---
+ arch/x86/mm/mmio-mod.c | 4 ++--
+ 1 file changed, 2 insertions(+), 2 deletions(-)
+
+diff --git a/arch/x86/mm/mmio-mod.c b/arch/x86/mm/mmio-mod.c
+index 0057a7accfb1..5448ad4d0703 100644
+--- a/arch/x86/mm/mmio-mod.c
++++ b/arch/x86/mm/mmio-mod.c
+@@ -385,7 +385,7 @@ static void enter_uniprocessor(void)
+       int cpu;
+       int err;
+-      if (downed_cpus == NULL &&
++      if (!cpumask_available(downed_cpus) &&
+           !alloc_cpumask_var(&downed_cpus, GFP_KERNEL)) {
+               pr_notice("Failed to allocate mask\n");
+               goto out;
+@@ -415,7 +415,7 @@ static void leave_uniprocessor(void)
+       int cpu;
+       int err;
+-      if (downed_cpus == NULL || cpumask_weight(downed_cpus) == 0)
++      if (!cpumask_available(downed_cpus) || cpumask_weight(downed_cpus) == 0)
+               return;
+       pr_notice("Re-enabling CPUs...\n");
+       for_each_cpu(cpu, downed_cpus) {
+-- 
+2.25.1
+