openwrt/target/linux/layerscape/patches-5.4/701-net-0236-enetc-Replace-enetc_gregs-with-a-readers-writer-lock.patch
Hauke Mehrtens 2464a9a8a4 kernel: bump 5.4 to 5.4.108
Refreshed all patches.

Compile-tested on: x86_64, ath79, lantiq
Runtime-tested on: x86_64, ath79

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
2021-03-27 16:26:56 +01:00

386 lines
12 KiB
Diff

From 659f899773f9f3fdc8325f61acc1017dd838126c Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <vladimir.oltean@nxp.com>
Date: Thu, 14 Nov 2019 17:30:50 +0200
Subject: [PATCH] enetc: Replace enetc_gregs with a readers-writer lock
The LS1028A MDIO errata tells us that any MDIO register access must not
be concurrent with any other ENETC register access.
That has been handled so far by a number of per-CPU spinlocks over the
ENETC register map. This came as an optimization over a single spinlock,
because the regular register accesses can still be concurrent with one
another, as long as they aren't concurrent with MDIO.
But this logic is broken in RT, because the enetc_rd_reg_wa and
enetc_wr_reg_wa functions can be preempted in any context, and when they
resume they may not run on the same CPU.
This renders the logic to take the per-CPU spinlock pointless, since the
spinlock may not be the correct one (corresponding to this CPU) after
preemption has occurred.
The following splat is telling us the same thing:
[ 19.073928] BUG: using smp_processor_id() in preemptible [00000000] code: systemd-network/3423
[ 19.073932] caller is debug_smp_processor_id+0x1c/0x30
[ 19.073935] CPU: 1 PID: 3423 Comm: systemd-network Not tainted 4.19.68-rt26 #1
[ 19.073936] Hardware name: LS1028A RDB Board (DT)
[ 19.073938] Call trace:
[ 19.073940] dump_backtrace+0x0/0x1a0
[ 19.073942] show_stack+0x24/0x30
[ 19.073945] dump_stack+0x9c/0xdc
[ 19.073948] check_preemption_disabled+0xe0/0x100
[ 19.073951] debug_smp_processor_id+0x1c/0x30
[ 19.073954] enetc_open+0x1b0/0xbc0
[ 19.073957] __dev_open+0xdc/0x160
[ 19.073960] __dev_change_flags+0x160/0x1d0
[ 19.073963] dev_change_flags+0x34/0x70
[ 19.073966] do_setlink+0x2a0/0xcd0
[ 19.073969] rtnl_setlink+0xe4/0x140
[ 19.073972] rtnetlink_rcv_msg+0x18c/0x500
[ 19.073975] netlink_rcv_skb+0x60/0x120
[ 19.073978] rtnetlink_rcv+0x28/0x40
[ 19.073982] netlink_unicast+0x194/0x210
[ 19.073985] netlink_sendmsg+0x194/0x330
[ 19.073987] sock_sendmsg+0x34/0x50
[ 19.073990] __sys_sendto+0xe4/0x150
[ 19.073992] __arm64_sys_sendto+0x30/0x40
[ 19.073996] el0_svc_common+0xa4/0x1a0
[ 19.073999] el0_svc_handler+0x38/0x80
[ 19.074002] el0_svc+0x8/0xc
But there already exists a spinlock optimized for the single writer,
multiple readers case: the rwlock_t. The writer in this case is the MDIO
access code (irrelevant whether that MDIO access is a register read or
write), and the reader is everybody else.
This patch also fixes two more existing bugs in the errata workaround:
- The MDIO access code was not unlocking the per-CPU spinlocks in the
reverse order of their locking order.
- The per-CPU spinlock array was not initialized.
Fixes: 5ec0d668d62e ("enetc: WA for MDIO register access issue")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Claudiu Manoil <claudiu.manoil@nxp.com>
---
drivers/net/ethernet/freescale/enetc/enetc.c | 59 ++++++----------------
drivers/net/ethernet/freescale/enetc/enetc_hw.h | 67 ++++---------------------
drivers/net/ethernet/freescale/enetc/enetc_pf.c | 5 +-
3 files changed, 30 insertions(+), 101 deletions(-)
--- a/drivers/net/ethernet/freescale/enetc/enetc.c
+++ b/drivers/net/ethernet/freescale/enetc/enetc.c
@@ -20,13 +20,8 @@ netdev_tx_t enetc_xmit(struct sk_buff *s
{
struct enetc_ndev_priv *priv = netdev_priv(ndev);
struct enetc_bdr *tx_ring;
- unsigned long flags;
- /* pointer to per-cpu ENETC lock for register access issue WA */
- spinlock_t *lock;
int count;
- lock = this_cpu_ptr(&enetc_gregs);
-
tx_ring = priv->tx_ring[skb->queue_mapping];
if (unlikely(skb_shinfo(skb)->nr_frags > ENETC_MAX_SKB_FRAGS))
@@ -39,11 +34,9 @@ netdev_tx_t enetc_xmit(struct sk_buff *s
return NETDEV_TX_BUSY;
}
- spin_lock_irqsave(lock, flags);
-
+ read_lock(&enetc_mdio_lock);
count = enetc_map_tx_buffs(tx_ring, skb, priv->active_offloads);
-
- spin_unlock_irqrestore(lock, flags);
+ read_unlock(&enetc_mdio_lock);
if (unlikely(!count))
goto drop_packet_err;
@@ -259,13 +252,9 @@ dma_err:
static irqreturn_t enetc_msix(int irq, void *data)
{
struct enetc_int_vector *v = data;
- unsigned long flags;
- /* pointer to per-cpu ENETC lock for register access issue WA */
- spinlock_t *lock;
int i;
- lock = this_cpu_ptr(&enetc_gregs);
- spin_lock_irqsave(lock, flags);
+ read_lock(&enetc_mdio_lock);
/* disable interrupts */
enetc_wr_reg_hot(v->rbier, 0);
@@ -273,7 +262,7 @@ static irqreturn_t enetc_msix(int irq, v
for_each_set_bit(i, &v->tx_rings_map, ENETC_MAX_NUM_TXQS)
enetc_wr_reg_hot(v->tbier_base + ENETC_BDR_OFF(i), 0);
- spin_unlock_irqrestore(lock, flags);
+ read_unlock(&enetc_mdio_lock);
napi_schedule_irqoff(&v->napi);
@@ -289,9 +278,6 @@ static int enetc_poll(struct napi_struct
struct enetc_int_vector
*v = container_of(napi, struct enetc_int_vector, napi);
bool complete = true;
- unsigned long flags;
- /* pointer to per-cpu ENETC lock for register access issue WA */
- spinlock_t *lock;
int work_done;
int i;
@@ -308,8 +294,7 @@ static int enetc_poll(struct napi_struct
napi_complete_done(napi, work_done);
- lock = this_cpu_ptr(&enetc_gregs);
- spin_lock_irqsave(lock, flags);
+ read_lock(&enetc_mdio_lock);
/* enable interrupts */
enetc_wr_reg_hot(v->rbier, ENETC_RBIER_RXTIE);
@@ -318,7 +303,7 @@ static int enetc_poll(struct napi_struct
enetc_wr_reg_hot(v->tbier_base + ENETC_BDR_OFF(i),
ENETC_TBIER_TXTIE);
- spin_unlock_irqrestore(lock, flags);
+ read_unlock(&enetc_mdio_lock);
return work_done;
}
@@ -363,18 +348,12 @@ static bool enetc_clean_tx_ring(struct e
bool do_tstamp;
u64 tstamp = 0;
- unsigned long flags;
- /* pointer to per-cpu ENETC lock for register access issue WA */
- spinlock_t *lock;
-
- lock = this_cpu_ptr(&enetc_gregs);
-
i = tx_ring->next_to_clean;
tx_swbd = &tx_ring->tx_swbd[i];
- spin_lock_irqsave(lock, flags);
+ read_lock(&enetc_mdio_lock);
bds_to_clean = enetc_bd_ready_count(tx_ring, i);
- spin_unlock_irqrestore(lock, flags);
+ read_unlock(&enetc_mdio_lock);
do_tstamp = false;
@@ -417,7 +396,7 @@ static bool enetc_clean_tx_ring(struct e
tx_swbd = tx_ring->tx_swbd;
}
- spin_lock_irqsave(lock, flags);
+ read_lock(&enetc_mdio_lock);
/* BD iteration loop end */
if (is_eof) {
@@ -430,7 +409,7 @@ static bool enetc_clean_tx_ring(struct e
if (unlikely(!bds_to_clean))
bds_to_clean = enetc_bd_ready_count(tx_ring, i);
- spin_unlock_irqrestore(lock, flags);
+ read_unlock(&enetc_mdio_lock);
}
tx_ring->next_to_clean = i;
@@ -516,7 +495,7 @@ static int enetc_refill_rx_ring(struct e
}
#ifdef CONFIG_FSL_ENETC_HW_TIMESTAMPING
-/* Must be called with &enetc_gregs spinlock held */
+/* Must be called with the read-side enetc_mdio_lock held */
static void enetc_get_rx_tstamp(struct net_device *ndev,
union enetc_rx_bd *rxbd,
struct sk_buff *skb)
@@ -667,12 +646,6 @@ static int enetc_clean_rx_ring(struct en
int rx_frm_cnt = 0, rx_byte_cnt = 0;
int cleaned_cnt, i;
- unsigned long flags;
- /* pointer to per-cpu ENETC lock for register access issue WA */
- spinlock_t *lock;
-
- lock = this_cpu_ptr(&enetc_gregs);
-
cleaned_cnt = enetc_bd_unused(rx_ring);
/* next descriptor to process */
i = rx_ring->next_to_clean;
@@ -683,7 +656,7 @@ static int enetc_clean_rx_ring(struct en
u32 bd_status;
u16 size;
- spin_lock_irqsave(lock, flags);
+ read_lock(&enetc_mdio_lock);
if (cleaned_cnt >= ENETC_RXBD_BUNDLE) {
int count = enetc_refill_rx_ring(rx_ring, cleaned_cnt);
@@ -694,7 +667,7 @@ static int enetc_clean_rx_ring(struct en
rxbd = ENETC_RXBD(*rx_ring, i);
bd_status = le32_to_cpu(rxbd->r.lstatus);
if (!bd_status) {
- spin_unlock_irqrestore(lock, flags);
+ read_unlock(&enetc_mdio_lock);
break;
}
@@ -703,7 +676,7 @@ static int enetc_clean_rx_ring(struct en
size = le16_to_cpu(rxbd->r.buf_len);
skb = enetc_map_rx_buff_to_skb(rx_ring, i, size);
if (!skb) {
- spin_unlock_irqrestore(lock, flags);
+ read_unlock(&enetc_mdio_lock);
break;
}
@@ -719,7 +692,7 @@ static int enetc_clean_rx_ring(struct en
if (unlikely(bd_status &
ENETC_RXBD_LSTATUS(ENETC_RXBD_ERR_MASK))) {
- spin_unlock_irqrestore(lock, flags);
+ read_unlock(&enetc_mdio_lock);
dev_kfree_skb(skb);
while (!(bd_status & ENETC_RXBD_LSTATUS_F)) {
dma_rmb();
@@ -763,7 +736,7 @@ static int enetc_clean_rx_ring(struct en
enetc_process_skb(rx_ring, skb);
- spin_unlock_irqrestore(lock, flags);
+ read_unlock(&enetc_mdio_lock);
napi_gro_receive(napi, skb);
--- a/drivers/net/ethernet/freescale/enetc/enetc_hw.h
+++ b/drivers/net/ethernet/freescale/enetc/enetc_hw.h
@@ -331,7 +331,7 @@ struct enetc_hw {
#define enetc_wr_reg(reg, val) enetc_wr_reg_wa((reg), (val))
/* accessors for data-path, due to MDIO issue on LS1028 these should be called
- * only under enetc_gregs per-cpu lock
+ * only under the rwlock_t enetc_mdio_lock
*/
#define enetc_rd_reg_hot(reg) ioread32((reg))
#define enetc_wr_reg_hot(reg, val) iowrite32((val), (reg))
@@ -354,90 +354,45 @@ static inline u64 enetc_rd_reg64(void __
}
#endif
-extern DEFINE_PER_CPU(spinlock_t, enetc_gregs);
+extern rwlock_t enetc_mdio_lock;
static inline u32 enetc_rd_reg_wa(void *reg)
{
- unsigned long flags;
- /* pointer to per-cpu ENETC lock for register access issue WA */
- spinlock_t *lock;
u32 val;
- lock = this_cpu_ptr(&enetc_gregs);
- spin_lock_irqsave(lock, flags);
+ read_lock(&enetc_mdio_lock);
val = ioread32(reg);
- spin_unlock_irqrestore(lock, flags);
+ read_unlock(&enetc_mdio_lock);
return val;
}
static inline void enetc_wr_reg_wa(void *reg, u32 val)
{
- unsigned long flags;
- /* pointer to per-cpu ENETC lock for register access issue WA */
- spinlock_t *lock;
-
- lock = this_cpu_ptr(&enetc_gregs);
- spin_lock_irqsave(lock, flags);
+ read_lock(&enetc_mdio_lock);
iowrite32(val, reg);
- spin_unlock_irqrestore(lock, flags);
+ read_unlock(&enetc_mdio_lock);
}
-/* NR_CPUS=256 in ARM64 defconfig and using it as array size triggers stack
- * frame warnings for the functions below. Use a custom define of 2 for now,
- * LS1028 has just two cores.
- */
-#define ENETC_NR_CPU_LOCKS 2
-
static inline u32 enetc_rd_reg_wa_single(void *reg)
{
- u32 val;
- int cpu;
- /* per-cpu ENETC lock array for register access issue WA */
- spinlock_t *lock[ENETC_NR_CPU_LOCKS];
unsigned long flags;
+ u32 val;
- local_irq_save(flags);
- preempt_disable();
-
- for_each_online_cpu(cpu) {
- lock[cpu] = per_cpu_ptr(&enetc_gregs, cpu);
- spin_lock(lock[cpu]);
- }
-
+ write_lock_irqsave(&enetc_mdio_lock, flags);
val = ioread32(reg);
-
- for_each_online_cpu(cpu)
- spin_unlock(lock[cpu]);
- local_irq_restore(flags);
-
- preempt_enable();
+ write_unlock_irqrestore(&enetc_mdio_lock, flags);
return val;
}
static inline void enetc_wr_reg_wa_single(void *reg, u32 val)
{
- int cpu;
- /* per-cpu ENETC lock array for register access issue WA */
- spinlock_t *lock[ENETC_NR_CPU_LOCKS];
unsigned long flags;
- local_irq_save(flags);
- preempt_disable();
-
- for_each_online_cpu(cpu) {
- lock[cpu] = per_cpu_ptr(&enetc_gregs, cpu);
- spin_lock(lock[cpu]);
- }
-
+ write_lock_irqsave(&enetc_mdio_lock, flags);
iowrite32(val, reg);
-
- for_each_online_cpu(cpu)
- spin_unlock(lock[cpu]);
- local_irq_restore(flags);
-
- preempt_enable();
+ write_unlock_irqrestore(&enetc_mdio_lock, flags);
}
#define enetc_rd(hw, off) enetc_rd_reg((hw)->reg + (off))
--- a/drivers/net/ethernet/freescale/enetc/enetc_pf.c
+++ b/drivers/net/ethernet/freescale/enetc/enetc_pf.c
@@ -1076,8 +1076,9 @@ static void enetc_pf_remove(struct pci_d
enetc_pci_remove(pdev);
}
-DEFINE_PER_CPU(spinlock_t, enetc_gregs);
-EXPORT_PER_CPU_SYMBOL(enetc_gregs);
+/* Lock for MDIO access errata on LS1028A */
+DEFINE_RWLOCK(enetc_mdio_lock);
+EXPORT_SYMBOL_GPL(enetc_mdio_lock);
static const struct pci_device_id enetc_pf_id_table[] = {
{ PCI_DEVICE(PCI_VENDOR_ID_FREESCALE, ENETC_DEV_ID_PF) },