More fixes for Raspberry Pi 3+ lan78xx ethernet interface, Fixes for Cavium ThunderX ZIP driver stability

This commit is contained in:
Peter Robinson 2018-04-09 18:30:51 +01:00 committed by Jeremy Cline
parent 197f0bb941
commit 33a8762bdc
No known key found for this signature in database
GPG Key ID: 9223308FA9B246DB
3 changed files with 765 additions and 0 deletions

View File

@ -0,0 +1,403 @@
From patchwork Mon Apr 9 15:45:50 2018
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: [v2,1/5] crypto: thunderx_zip: Fix fallout from CONFIG_VMAP_STACK
From: Jan Glauber <jglauber@cavium.com>
X-Patchwork-Id: 10331719
Message-Id: <20180409154554.7578-2-jglauber@cavium.com>
To: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "David S . Miller" <davem@davemloft.net>,
linux-crypto@vger.kernel.org, linux-kernel@vger.kernel.org,
Mahipal Challa <Mahipal.Challa@cavium.com>,
Robert Richter <rrichter@cavium.com>, Jan Glauber <jglauber@cavium.com>,
stable <stable@vger.kernel.org>
Date: Mon, 9 Apr 2018 17:45:50 +0200
Enabling virtual mapped kernel stacks breaks the thunderx_zip
driver. On compression or decompression the executing CPU hangs
in an endless loop. The reason for this is the usage of __pa
by the driver which does no longer work for an address that is
not part of the 1:1 mapping.
The zip driver allocates a result struct on the stack and needs
to tell the hardware the physical address within this struct
that is used to signal the completion of the request.
As the hardware gets the wrong address after the broken __pa
conversion it writes to an arbitrary address. The zip driver then
waits forever for the completion byte to contain a non-zero value.
Allocating the result struct from 1:1 mapped memory resolves this
bug.
Signed-off-by: Jan Glauber <jglauber@cavium.com>
Reviewed-by: Robert Richter <rrichter@cavium.com>
Cc: stable <stable@vger.kernel.org> # 4.14
---
drivers/crypto/cavium/zip/zip_crypto.c | 22 ++++++++++++++--------
1 file changed, 14 insertions(+), 8 deletions(-)
diff --git a/drivers/crypto/cavium/zip/zip_crypto.c b/drivers/crypto/cavium/zip/zip_crypto.c
index 8df4d26cf9d4..b92b6e7e100f 100644
--- a/drivers/crypto/cavium/zip/zip_crypto.c
+++ b/drivers/crypto/cavium/zip/zip_crypto.c
@@ -124,7 +124,7 @@ int zip_compress(const u8 *src, unsigned int slen,
struct zip_kernel_ctx *zip_ctx)
{
struct zip_operation *zip_ops = NULL;
- struct zip_state zip_state;
+ struct zip_state *zip_state;
struct zip_device *zip = NULL;
int ret;
@@ -135,20 +135,23 @@ int zip_compress(const u8 *src, unsigned int slen,
if (!zip)
return -ENODEV;
- memset(&zip_state, 0, sizeof(struct zip_state));
+ zip_state = kzalloc(sizeof(*zip_state), GFP_ATOMIC);
+ if (!zip_state)
+ return -ENOMEM;
+
zip_ops = &zip_ctx->zip_comp;
zip_ops->input_len = slen;
zip_ops->output_len = *dlen;
memcpy(zip_ops->input, src, slen);
- ret = zip_deflate(zip_ops, &zip_state, zip);
+ ret = zip_deflate(zip_ops, zip_state, zip);
if (!ret) {
*dlen = zip_ops->output_len;
memcpy(dst, zip_ops->output, *dlen);
}
-
+ kfree(zip_state);
return ret;
}
@@ -157,7 +160,7 @@ int zip_decompress(const u8 *src, unsigned int slen,
struct zip_kernel_ctx *zip_ctx)
{
struct zip_operation *zip_ops = NULL;
- struct zip_state zip_state;
+ struct zip_state *zip_state;
struct zip_device *zip = NULL;
int ret;
@@ -168,7 +171,10 @@ int zip_decompress(const u8 *src, unsigned int slen,
if (!zip)
return -ENODEV;
- memset(&zip_state, 0, sizeof(struct zip_state));
+ zip_state = kzalloc(sizeof(*zip_state), GFP_ATOMIC);
+ if (!zip_state)
+ return -ENOMEM;
+
zip_ops = &zip_ctx->zip_decomp;
memcpy(zip_ops->input, src, slen);
@@ -179,13 +185,13 @@ int zip_decompress(const u8 *src, unsigned int slen,
zip_ops->input_len = slen;
zip_ops->output_len = *dlen;
- ret = zip_inflate(zip_ops, &zip_state, zip);
+ ret = zip_inflate(zip_ops, zip_state, zip);
if (!ret) {
*dlen = zip_ops->output_len;
memcpy(dst, zip_ops->output, *dlen);
}
-
+ kfree(zip_state);
return ret;
}
From patchwork Mon Apr 9 15:45:51 2018
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: [v2,2/5] crypto: thunderx_zip: Limit result reading attempts
From: Jan Glauber <jglauber@cavium.com>
X-Patchwork-Id: 10331705
Message-Id: <20180409154554.7578-3-jglauber@cavium.com>
To: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "David S . Miller" <davem@davemloft.net>,
linux-crypto@vger.kernel.org, linux-kernel@vger.kernel.org,
Mahipal Challa <Mahipal.Challa@cavium.com>,
Robert Richter <rrichter@cavium.com>, Jan Glauber <jglauber@cavium.com>,
stable <stable@vger.kernel.org>
Date: Mon, 9 Apr 2018 17:45:51 +0200
After issuing a request an endless loop was used to read the
completion state from memory which is asynchronously updated
by the ZIP coprocessor.
Add an upper bound to the retry attempts to prevent a CPU getting stuck
forever in case of an error. Additionally, add a read memory barrier
and a small delay between the reading attempts.
Signed-off-by: Jan Glauber <jglauber@cavium.com>
Reviewed-by: Robert Richter <rrichter@cavium.com>
Cc: stable <stable@vger.kernel.org> # 4.14
---
drivers/crypto/cavium/zip/common.h | 21 +++++++++++++++++++++
drivers/crypto/cavium/zip/zip_deflate.c | 4 ++--
drivers/crypto/cavium/zip/zip_inflate.c | 4 ++--
3 files changed, 25 insertions(+), 4 deletions(-)
diff --git a/drivers/crypto/cavium/zip/common.h b/drivers/crypto/cavium/zip/common.h
index dc451e0a43c5..58fb3ed6e644 100644
--- a/drivers/crypto/cavium/zip/common.h
+++ b/drivers/crypto/cavium/zip/common.h
@@ -46,8 +46,10 @@
#ifndef __COMMON_H__
#define __COMMON_H__
+#include <linux/delay.h>
#include <linux/init.h>
#include <linux/interrupt.h>
+#include <linux/io.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/pci.h>
@@ -149,6 +151,25 @@ struct zip_operation {
u32 sizeofzops;
};
+static inline int zip_poll_result(union zip_zres_s *result)
+{
+ int retries = 1000;
+
+ while (!result->s.compcode) {
+ if (!--retries) {
+ pr_err("ZIP ERR: request timed out");
+ return -ETIMEDOUT;
+ }
+ udelay(10);
+ /*
+ * Force re-reading of compcode which is updated
+ * by the ZIP coprocessor.
+ */
+ rmb();
+ }
+ return 0;
+}
+
/* error messages */
#define zip_err(fmt, args...) pr_err("ZIP ERR:%s():%d: " \
fmt "\n", __func__, __LINE__, ## args)
diff --git a/drivers/crypto/cavium/zip/zip_deflate.c b/drivers/crypto/cavium/zip/zip_deflate.c
index 9a944b8c1e29..d7133f857d67 100644
--- a/drivers/crypto/cavium/zip/zip_deflate.c
+++ b/drivers/crypto/cavium/zip/zip_deflate.c
@@ -129,8 +129,8 @@ int zip_deflate(struct zip_operation *zip_ops, struct zip_state *s,
/* Stats update for compression requests submitted */
atomic64_inc(&zip_dev->stats.comp_req_submit);
- while (!result_ptr->s.compcode)
- continue;
+ /* Wait for completion or error */
+ zip_poll_result(result_ptr);
/* Stats update for compression requests completed */
atomic64_inc(&zip_dev->stats.comp_req_complete);
diff --git a/drivers/crypto/cavium/zip/zip_inflate.c b/drivers/crypto/cavium/zip/zip_inflate.c
index 50cbdd83dbf2..7e0d73e2f89e 100644
--- a/drivers/crypto/cavium/zip/zip_inflate.c
+++ b/drivers/crypto/cavium/zip/zip_inflate.c
@@ -143,8 +143,8 @@ int zip_inflate(struct zip_operation *zip_ops, struct zip_state *s,
/* Decompression requests submitted stats update */
atomic64_inc(&zip_dev->stats.decomp_req_submit);
- while (!result_ptr->s.compcode)
- continue;
+ /* Wait for completion or error */
+ zip_poll_result(result_ptr);
/* Decompression requests completed stats update */
atomic64_inc(&zip_dev->stats.decomp_req_complete);
From patchwork Mon Apr 9 15:45:52 2018
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: [v2,3/5] crypto: thunderx_zip: Prevent division by zero
From: Jan Glauber <jglauber@cavium.com>
X-Patchwork-Id: 10331709
Message-Id: <20180409154554.7578-4-jglauber@cavium.com>
To: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "David S . Miller" <davem@davemloft.net>,
linux-crypto@vger.kernel.org, linux-kernel@vger.kernel.org,
Mahipal Challa <Mahipal.Challa@cavium.com>,
Robert Richter <rrichter@cavium.com>, Jan Glauber <jglauber@cavium.com>
Date: Mon, 9 Apr 2018 17:45:52 +0200
Avoid two potential divisions by zero when calculating average
values for the zip statistics.
Signed-off-by: Jan Glauber <jglauber@cavium.com>
Reviewed-by: Robert Richter <rrichter@cavium.com>
---
drivers/crypto/cavium/zip/zip_main.c | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/drivers/crypto/cavium/zip/zip_main.c b/drivers/crypto/cavium/zip/zip_main.c
index 1cd8aa488185..79b449e0f955 100644
--- a/drivers/crypto/cavium/zip/zip_main.c
+++ b/drivers/crypto/cavium/zip/zip_main.c
@@ -482,10 +482,11 @@ static int zip_show_stats(struct seq_file *s, void *unused)
atomic64_add(val, &st->pending_req);
}
- avg_chunk = (atomic64_read(&st->comp_in_bytes) /
- atomic64_read(&st->comp_req_complete));
- avg_cr = (atomic64_read(&st->comp_in_bytes) /
- atomic64_read(&st->comp_out_bytes));
+ val = atomic64_read(&st->comp_req_complete);
+ avg_chunk = (val) ? atomic64_read(&st->comp_in_bytes) / val : 0;
+
+ val = atomic64_read(&st->comp_out_bytes);
+ avg_cr = (val) ? atomic64_read(&st->comp_in_bytes) / val : 0;
seq_printf(s, " ZIP Device %d Stats\n"
"-----------------------------------\n"
"Comp Req Submitted : \t%lld\n"
From patchwork Mon Apr 9 15:45:53 2018
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: [v2,4/5] crypto: thunderx_zip: Fix statistics pending request value
From: Jan Glauber <jglauber@cavium.com>
X-Patchwork-Id: 10331711
Message-Id: <20180409154554.7578-5-jglauber@cavium.com>
To: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "David S . Miller" <davem@davemloft.net>,
linux-crypto@vger.kernel.org, linux-kernel@vger.kernel.org,
Mahipal Challa <Mahipal.Challa@cavium.com>,
Robert Richter <rrichter@cavium.com>, Jan Glauber <jglauber@cavium.com>
Date: Mon, 9 Apr 2018 17:45:53 +0200
The pending request counter was read from the wrong register. While
at it, there is no need to use an atomic for it as it is only read
localy in a loop.
Signed-off-by: Jan Glauber <jglauber@cavium.com>
Reviewed-by: Robert Richter <rrichter@cavium.com>
---
drivers/crypto/cavium/zip/zip_main.c | 13 +++++--------
drivers/crypto/cavium/zip/zip_main.h | 1 -
2 files changed, 5 insertions(+), 9 deletions(-)
diff --git a/drivers/crypto/cavium/zip/zip_main.c b/drivers/crypto/cavium/zip/zip_main.c
index 79b449e0f955..ae5b20c695ca 100644
--- a/drivers/crypto/cavium/zip/zip_main.c
+++ b/drivers/crypto/cavium/zip/zip_main.c
@@ -469,6 +469,8 @@ static int zip_show_stats(struct seq_file *s, void *unused)
struct zip_stats *st;
for (index = 0; index < MAX_ZIP_DEVICES; index++) {
+ u64 pending = 0;
+
if (zip_dev[index]) {
zip = zip_dev[index];
st = &zip->stats;
@@ -476,10 +478,8 @@ static int zip_show_stats(struct seq_file *s, void *unused)
/* Get all the pending requests */
for (q = 0; q < ZIP_NUM_QUEUES; q++) {
val = zip_reg_read((zip->reg_base +
- ZIP_DBG_COREX_STA(q)));
- val = (val >> 32);
- val = val & 0xffffff;
- atomic64_add(val, &st->pending_req);
+ ZIP_DBG_QUEX_STA(q)));
+ pending += val >> 32 & 0xffffff;
}
val = atomic64_read(&st->comp_req_complete);
@@ -514,10 +514,7 @@ static int zip_show_stats(struct seq_file *s, void *unused)
(u64)atomic64_read(&st->decomp_in_bytes),
(u64)atomic64_read(&st->decomp_out_bytes),
(u64)atomic64_read(&st->decomp_bad_reqs),
- (u64)atomic64_read(&st->pending_req));
-
- /* Reset pending requests count */
- atomic64_set(&st->pending_req, 0);
+ pending);
}
}
return 0;
diff --git a/drivers/crypto/cavium/zip/zip_main.h b/drivers/crypto/cavium/zip/zip_main.h
index 64e051f60784..e1e4fa92ce80 100644
--- a/drivers/crypto/cavium/zip/zip_main.h
+++ b/drivers/crypto/cavium/zip/zip_main.h
@@ -74,7 +74,6 @@ struct zip_stats {
atomic64_t comp_req_complete;
atomic64_t decomp_req_submit;
atomic64_t decomp_req_complete;
- atomic64_t pending_req;
atomic64_t comp_in_bytes;
atomic64_t comp_out_bytes;
atomic64_t decomp_in_bytes;
From patchwork Mon Apr 9 15:45:54 2018
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: [v2,5/5] crypto: thunderx_zip: Fix smp_processor_id() warnings
From: Jan Glauber <jglauber@cavium.com>
X-Patchwork-Id: 10331715
Message-Id: <20180409154554.7578-6-jglauber@cavium.com>
To: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "David S . Miller" <davem@davemloft.net>,
linux-crypto@vger.kernel.org, linux-kernel@vger.kernel.org,
Mahipal Challa <Mahipal.Challa@cavium.com>,
Robert Richter <rrichter@cavium.com>, Jan Glauber <jglauber@cavium.com>
Date: Mon, 9 Apr 2018 17:45:54 +0200
Switch to raw_smp_processor_id() to prevent a number of
warnings from kernel debugging. We do not care about
preemption here, as the CPU number is only used as a
poor mans load balancing or device selection. If preemption
happens during a compress/decompress operation a small performance
hit will occur but everything will continue to work, so just
ignore it.
Signed-off-by: Jan Glauber <jglauber@cavium.com>
Reviewed-by: Robert Richter <rrichter@cavium.com>
---
drivers/crypto/cavium/zip/zip_device.c | 4 ++--
drivers/crypto/cavium/zip/zip_main.c | 2 +-
2 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/crypto/cavium/zip/zip_device.c b/drivers/crypto/cavium/zip/zip_device.c
index ccf21fb91513..f174ec29ed69 100644
--- a/drivers/crypto/cavium/zip/zip_device.c
+++ b/drivers/crypto/cavium/zip/zip_device.c
@@ -87,12 +87,12 @@ u32 zip_load_instr(union zip_inst_s *instr,
* Distribute the instructions between the enabled queues based on
* the CPU id.
*/
- if (smp_processor_id() % 2 == 0)
+ if (raw_smp_processor_id() % 2 == 0)
queue = 0;
else
queue = 1;
- zip_dbg("CPU Core: %d Queue number:%d", smp_processor_id(), queue);
+ zip_dbg("CPU Core: %d Queue number:%d", raw_smp_processor_id(), queue);
/* Take cmd buffer lock */
spin_lock(&zip_dev->iq[queue].lock);
diff --git a/drivers/crypto/cavium/zip/zip_main.c b/drivers/crypto/cavium/zip/zip_main.c
index ae5b20c695ca..be055b9547f6 100644
--- a/drivers/crypto/cavium/zip/zip_main.c
+++ b/drivers/crypto/cavium/zip/zip_main.c
@@ -113,7 +113,7 @@ struct zip_device *zip_get_device(int node)
*/
int zip_get_node_id(void)
{
- return cpu_to_node(smp_processor_id());
+ return cpu_to_node(raw_smp_processor_id());
}
/* Initializes the ZIP h/w sub-system */

View File

@ -106,3 +106,358 @@ index 60a604cc7647..a21039852f8d 100644
addr_lo = addr[0] | (addr[1] << 8) |
(addr[2] << 16) | (addr[3] << 24);
addr_hi = addr[4] | (addr[5] << 8);
From b5284e5d2d3562dac311443969a538b7fecb9848 Mon Sep 17 00:00:00 2001
From: Phil Elwell <phil@raspberrypi.org>
Date: Wed, 28 Mar 2018 12:18:13 +0100
Subject: [PATCH 1/5] lan78xx: Ignore DT MAC address if already valid
The patch to set the lan78xx MAC address from DT does so regardless of
whether or not the interface already has a valid address. As the
initialisation function is called from the reset handler when the
interface is brought up, it is impossible to change the MAC address
in a way that persists across the interface being brought up.
Fix the problem by moving the DT reading code after the check for a
valid address.
See: https://www.raspberrypi.org/forums/viewtopic.php?f=28&t=209309
Signed-off-by: Phil Elwell <phil@raspberrypi.org>
---
drivers/net/usb/lan78xx.c | 17 +++++++++--------
1 file changed, 9 insertions(+), 8 deletions(-)
diff --git a/drivers/net/usb/lan78xx.c b/drivers/net/usb/lan78xx.c
index b43b16b6e7ee..97ee7d3f749d 100644
--- a/drivers/net/usb/lan78xx.c
+++ b/drivers/net/usb/lan78xx.c
@@ -1641,14 +1641,6 @@ static void lan78xx_init_mac_address(struct lan78xx_net *dev)
u32 addr_lo, addr_hi;
int ret;
u8 addr[6];
- const u8 *mac_addr;
-
- /* maybe the boot loader passed the MAC address in devicetree */
- mac_addr = of_get_mac_address(dev->udev->dev.of_node);
- if (mac_addr) {
- ether_addr_copy(addr, mac_addr);
- goto set_mac_addr;
- }
ret = lan78xx_read_reg(dev, RX_ADDRL, &addr_lo);
ret = lan78xx_read_reg(dev, RX_ADDRH, &addr_hi);
@@ -1661,6 +1653,15 @@ static void lan78xx_init_mac_address(struct lan78xx_net *dev)
addr[5] = (addr_hi >> 8) & 0xFF;
if (!is_valid_ether_addr(addr)) {
+ const u8 *mac_addr;
+
+ /* maybe the boot loader passed the MAC address in devicetree */
+ mac_addr = of_get_mac_address(dev->udev->dev.of_node);
+ if (mac_addr) {
+ ether_addr_copy(addr, mac_addr);
+ goto set_mac_addr;
+ }
+
/* reading mac address from EEPROM or OTP */
if ((lan78xx_read_eeprom(dev, EEPROM_MAC_OFFSET, ETH_ALEN,
addr) == 0) ||
--
2.17.0
From 2c5d6ac9133cbfed05b97b34246121bddaf2aea4 Mon Sep 17 00:00:00 2001
From: Dave Stevenson <dave.stevenson@raspberrypi.org>
Date: Wed, 4 Apr 2018 16:34:24 +0100
Subject: [PATCH 2/5] net: lan78xx: Allow for VLAN headers in timeout.
The frame abort timeout being set by lan78xx_set_rx_max_frame_length
didn't account for any VLAN headers, resulting in very low
throughput if used with tagged VLANs.
Use VLAN_ETH_HLEN instead of ETH_HLEN to correct for this.
See https://github.com/raspberrypi/linux/issues/2458
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
---
drivers/net/usb/lan78xx.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/drivers/net/usb/lan78xx.c b/drivers/net/usb/lan78xx.c
index 97ee7d3f749d..5fd7b8569cba 100644
--- a/drivers/net/usb/lan78xx.c
+++ b/drivers/net/usb/lan78xx.c
@@ -2193,7 +2193,7 @@ static int lan78xx_change_mtu(struct net_device *netdev, int new_mtu)
if ((ll_mtu % dev->maxpacket) == 0)
return -EDOM;
- ret = lan78xx_set_rx_max_frame_length(dev, new_mtu + ETH_HLEN);
+ ret = lan78xx_set_rx_max_frame_length(dev, new_mtu + VLAN_ETH_HLEN);
netdev->mtu = new_mtu;
@@ -2488,7 +2488,8 @@ static int lan78xx_reset(struct lan78xx_net *dev)
buf |= FCT_TX_CTL_EN_;
ret = lan78xx_write_reg(dev, FCT_TX_CTL, buf);
- ret = lan78xx_set_rx_max_frame_length(dev, dev->net->mtu + ETH_HLEN);
+ ret = lan78xx_set_rx_max_frame_length(dev,
+ dev->net->mtu + VLAN_ETH_HLEN);
ret = lan78xx_read_reg(dev, MAC_RX, &buf);
buf |= MAC_RX_RXEN_;
--
2.17.0
From 833315351413d94d7db407847448dfeddfafe127 Mon Sep 17 00:00:00 2001
From: Peter Robinson <pbrobinson@gmail.com>
Date: Mon, 9 Apr 2018 17:51:35 +0100
Subject: [PATCH 3/5] lan78xx: Connect phy early
When using wicked with a lan78xx device attached to the system, we
end up with ethtool commands issued on the device before an ifup
got issued. That lead to the following crash:
Unable to handle kernel NULL pointer dereference at virtual address 0000039c
pgd = ffff800035b30000
[0000039c] *pgd=0000000000000000
Internal error: Oops: 96000004 [#1] SMP
Modules linked in: [...]
Supported: Yes
CPU: 3 PID: 638 Comm: wickedd Tainted: G E 4.12.14-0-default #1
Hardware name: raspberrypi rpi/rpi, BIOS 2018.03-rc2 02/21/2018
task: ffff800035e74180 task.stack: ffff800036718000
PC is at phy_ethtool_ksettings_get+0x20/0x98
LR is at lan78xx_get_link_ksettings+0x44/0x60 [lan78xx]
pc : [<ffff0000086f7f30>] lr : [<ffff000000dcca84>] pstate: 20000005
sp : ffff80003671bb20
x29: ffff80003671bb20 x28: ffff800035e74180
x27: ffff000008912000 x26: 000000000000001d
x25: 0000000000000124 x24: ffff000008f74d00
x23: 0000004000114809 x22: 0000000000000000
x21: ffff80003671bbd0 x20: 0000000000000000
x19: ffff80003671bbd0 x18: 000000000000040d
x17: 0000000000000001 x16: 0000000000000000
x15: 0000000000000000 x14: ffffffffffffffff
x13: 0000000000000000 x12: 0000000000000020
x11: 0101010101010101 x10: fefefefefefefeff
x9 : 7f7f7f7f7f7f7f7f x8 : fefefeff31677364
x7 : 0000000080808080 x6 : ffff80003671bc9c
x5 : ffff80003671b9f8 x4 : ffff80002c296190
x3 : 0000000000000000 x2 : 0000000000000000
x1 : ffff80003671bbd0 x0 : ffff80003671bc00
Process wickedd (pid: 638, stack limit = 0xffff800036718000)
Call trace:
Exception stack(0xffff80003671b9e0 to 0xffff80003671bb20)
b9e0: ffff80003671bc00 ffff80003671bbd0 0000000000000000 0000000000000000
ba00: ffff80002c296190 ffff80003671b9f8 ffff80003671bc9c 0000000080808080
ba20: fefefeff31677364 7f7f7f7f7f7f7f7f fefefefefefefeff 0101010101010101
ba40: 0000000000000020 0000000000000000 ffffffffffffffff 0000000000000000
ba60: 0000000000000000 0000000000000001 000000000000040d ffff80003671bbd0
ba80: 0000000000000000 ffff80003671bbd0 0000000000000000 0000004000114809
baa0: ffff000008f74d00 0000000000000124 000000000000001d ffff000008912000
bac0: ffff800035e74180 ffff80003671bb20 ffff000000dcca84 ffff80003671bb20
bae0: ffff0000086f7f30 0000000020000005 ffff80002c296000 ffff800035223900
bb00: 0000ffffffffffff 0000000000000000 ffff80003671bb20 ffff0000086f7f30
[<ffff0000086f7f30>] phy_ethtool_ksettings_get+0x20/0x98
[<ffff000000dcca84>] lan78xx_get_link_ksettings+0x44/0x60 [lan78xx]
[<ffff0000087cbc40>] ethtool_get_settings+0x68/0x210
[<ffff0000087cc0d4>] dev_ethtool+0x214/0x2180
[<ffff0000087e5008>] dev_ioctl+0x400/0x630
[<ffff00000879dd00>] sock_do_ioctl+0x70/0x88
[<ffff00000879f5f8>] sock_ioctl+0x208/0x368
[<ffff0000082cde10>] do_vfs_ioctl+0xb0/0x848
[<ffff0000082ce634>] SyS_ioctl+0x8c/0xa8
Exception stack(0xffff80003671bec0 to 0xffff80003671c000)
bec0: 0000000000000009 0000000000008946 0000fffff4e841d0 0000aa0032687465
bee0: 0000aaaafa2319d4 0000fffff4e841d4 0000000032687465 0000000032687465
bf00: 000000000000001d 7f7fff7f7f7f7f7f 72606b622e71ff4c 7f7f7f7f7f7f7f7f
bf20: 0101010101010101 0000000000000020 ffffffffffffffff 0000ffff7f510c68
bf40: 0000ffff7f6a9d18 0000ffff7f44ce30 000000000000040d 0000ffff7f6f98f0
bf60: 0000fffff4e842c0 0000000000000001 0000aaaafa2c2e00 0000ffff7f6ab000
bf80: 0000fffff4e842c0 0000ffff7f62a000 0000aaaafa2b9f20 0000aaaafa2c2e00
bfa0: 0000fffff4e84818 0000fffff4e841a0 0000ffff7f5ad0cc 0000fffff4e841a0
bfc0: 0000ffff7f44ce3c 0000000080000000 0000000000000009 000000000000001d
bfe0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
The culprit is quite simple: The driver tries to access the phy left and right,
but only actually has a working reference to it when the device is up.
The fix thus is quite simple too: Get a reference to the phy on probe already
and keep it even when the device is going down.
With this patch applied, I can successfully run wicked on my system and bring
the interface up and down as many times as I want, without getting NULL pointer
dereferences in between.
Signed-off-by: Alexander Graf <agraf@suse.de>
---
drivers/net/usb/lan78xx.c | 34 ++++++++++++++++++----------------
1 file changed, 18 insertions(+), 16 deletions(-)
diff --git a/drivers/net/usb/lan78xx.c b/drivers/net/usb/lan78xx.c
index 5fd7b8569cba..60fa1257721c 100644
--- a/drivers/net/usb/lan78xx.c
+++ b/drivers/net/usb/lan78xx.c
@@ -2094,10 +2094,6 @@ static int lan78xx_phy_init(struct lan78xx_net *dev)
dev->fc_autoneg = phydev->autoneg;
- phy_start(phydev);
-
- netif_dbg(dev, ifup, dev->net, "phy initialised successfully");
-
return 0;
error:
@@ -2541,9 +2537,9 @@ static int lan78xx_open(struct net_device *net)
if (ret < 0)
goto done;
- ret = lan78xx_phy_init(dev);
- if (ret < 0)
- goto done;
+ phy_start(net->phydev);
+
+ netif_dbg(dev, ifup, dev->net, "phy initialised successfully");
/* for Link Check */
if (dev->urb_intr) {
@@ -2604,13 +2600,8 @@ static int lan78xx_stop(struct net_device *net)
if (timer_pending(&dev->stat_monitor))
del_timer_sync(&dev->stat_monitor);
- phy_unregister_fixup_for_uid(PHY_KSZ9031RNX, 0xfffffff0);
- phy_unregister_fixup_for_uid(PHY_LAN8835, 0xfffffff0);
-
- phy_stop(net->phydev);
- phy_disconnect(net->phydev);
-
- net->phydev = NULL;
+ if (net->phydev)
+ phy_stop(net->phydev);
clear_bit(EVENT_DEV_OPEN, &dev->flags);
netif_stop_queue(net);
@@ -3525,8 +3516,13 @@ static void lan78xx_disconnect(struct usb_interface *intf)
return;
udev = interface_to_usbdev(intf);
-
net = dev->net;
+
+ phy_unregister_fixup_for_uid(PHY_KSZ9031RNX, 0xfffffff0);
+ phy_unregister_fixup_for_uid(PHY_LAN8835, 0xfffffff0);
+
+ phy_disconnect(net->phydev);
+
unregister_netdev(net);
cancel_delayed_work_sync(&dev->wq);
@@ -3682,8 +3678,14 @@ static int lan78xx_probe(struct usb_interface *intf,
pm_runtime_set_autosuspend_delay(&udev->dev,
DEFAULT_AUTOSUSPEND_DELAY);
+ ret = lan78xx_phy_init(dev);
+ if (ret < 0)
+ goto out4;
+
return 0;
+out4:
+ unregister_netdev(netdev);
out3:
lan78xx_unbind(dev, intf);
out2:
@@ -4031,7 +4033,7 @@ static int lan78xx_reset_resume(struct usb_interface *intf)
lan78xx_reset(dev);
- lan78xx_phy_init(dev);
+ phy_start(dev->net->phydev);
return lan78xx_resume(intf);
}
--
2.17.0
From 7b4cc4a0af02c0d798007a143efa7509711d52d7 Mon Sep 17 00:00:00 2001
From: Phil Elwell <phil@raspberrypi.org>
Date: Wed, 4 Apr 2018 16:39:44 +0100
Subject: [PATCH 4/5] lan78xx: Don't reset the interface on open
With Alexander Graf's patch ("lan78xx: Connect phy early") applied,
the call to lan78xx_reset within lan78xx_open prevents the phy
interrupt from being generated (even though the link is up).
Avoid this issue by removing the lan78xx_reset call.
See: https://github.com/raspberrypi/linux/issues/2437
https://github.com/raspberrypi/linux/issues/2442
https://github.com/raspberrypi/linux/issues/2457
---
drivers/net/usb/lan78xx.c | 4 ----
1 file changed, 4 deletions(-)
diff --git a/drivers/net/usb/lan78xx.c b/drivers/net/usb/lan78xx.c
index 60fa1257721c..293ed1847932 100644
--- a/drivers/net/usb/lan78xx.c
+++ b/drivers/net/usb/lan78xx.c
@@ -2533,10 +2533,6 @@ static int lan78xx_open(struct net_device *net)
if (ret < 0)
goto out;
- ret = lan78xx_reset(dev);
- if (ret < 0)
- goto done;
-
phy_start(net->phydev);
netif_dbg(dev, ifup, dev->net, "phy initialised successfully");
--
2.17.0
From ddbd11509f01c388b968872aeabf630654275b0a Mon Sep 17 00:00:00 2001
From: Dave Stevenson <dave.stevenson@raspberrypi.org>
Date: Mon, 9 Apr 2018 14:31:54 +0100
Subject: [PATCH 5/5] net: lan78xx: Request s/w csum check on VLAN tagged
packets.
There appears to be some issue in the LAN78xx where the checksum
computed on a VLAN tagged packet is incorrect, or at least not
in the form that the kernel is after. This is most easily shown
by pinging a device via a VLAN tagged interface and it will dump
out the error message and stack trace from netdev_rx_csum_fault.
It has also been seen with standard TCP and UDP packets.
Until this is fully understood, request that the network stack
computes the checksum on packets signalled as having a VLAN tag
applied.
See https://github.com/raspberrypi/linux/issues/2458
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
---
drivers/net/usb/lan78xx.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/drivers/net/usb/lan78xx.c b/drivers/net/usb/lan78xx.c
index 293ed1847932..44cabda17bb6 100644
--- a/drivers/net/usb/lan78xx.c
+++ b/drivers/net/usb/lan78xx.c
@@ -2937,8 +2937,12 @@ static void lan78xx_rx_csum_offload(struct lan78xx_net *dev,
struct sk_buff *skb,
u32 rx_cmd_a, u32 rx_cmd_b)
{
+ /* Checksum offload appears to be flawed if used with VLANs.
+ * Elect for sw checksum check instead.
+ */
if (!(dev->net->features & NETIF_F_RXCSUM) ||
- unlikely(rx_cmd_a & RX_CMD_A_ICSM_)) {
+ unlikely(rx_cmd_a & RX_CMD_A_ICSM_) ||
+ (rx_cmd_a & RX_CMD_A_FVTG_)) {
skb->ip_summed = CHECKSUM_NONE;
} else {
skb->csum = ntohs((u16)(rx_cmd_b >> RX_CMD_B_CSUM_SHIFT_));
--
2.17.0

View File

@ -605,6 +605,9 @@ Patch311: bcm2835-hwrng-Handle-deferred-clock-properly.patch
Patch312: bcm283x-clk-audio-fixes.patch
# https://marc.info/?l=linux-kernel&m=152328880417846&w=2
Patch313: arm64-thunderx-crypto-zip-fixes.patch
# Enabling Patches for the RPi3+
Patch320: bcm2837-rpi-initial-support-for-the-3.patch
Patch321: bcm2837-gpio-expander.patch
@ -1879,6 +1882,10 @@ fi
#
#
%changelog
* Mon Apr 9 2018 Peter Robinson <pbrobinson@fedoraproject.org>
- More fixes for Raspberry Pi 3+ lan78xx ethernet interface
- Fixes for Cavium ThunderX ZIP driver stability
* Mon Apr 09 2018 Jeremy Cline <jeremy@jcline.org> - 4.16.1-200
- Linux v4.16.1