linux-stable/drivers/spmi/spmi-pmic-arb.c

1678 lines
46 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
/*
* Copyright (c) 2012-2015, 2017, 2021, The Linux Foundation. All rights reserved.
*/
#include <linux/bitmap.h>
#include <linux/delay.h>
#include <linux/err.h>
#include <linux/interrupt.h>
#include <linux/io.h>
#include <linux/irqchip/chained_irq.h>
#include <linux/irqdomain.h>
#include <linux/irq.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/of.h>
#include <linux/platform_device.h>
#include <linux/slab.h>
#include <linux/spmi.h>
/* PMIC Arbiter configuration registers */
#define PMIC_ARB_VERSION 0x0000
#define PMIC_ARB_VERSION_V2_MIN 0x20010000
#define PMIC_ARB_VERSION_V3_MIN 0x30000000
#define PMIC_ARB_VERSION_V5_MIN 0x50000000
#define PMIC_ARB_VERSION_V7_MIN 0x70000000
#define PMIC_ARB_INT_EN 0x0004
#define PMIC_ARB_FEATURES 0x0004
#define PMIC_ARB_FEATURES_PERIPH_MASK GENMASK(10, 0)
#define PMIC_ARB_FEATURES1 0x0008
/* PMIC Arbiter channel registers offsets */
#define PMIC_ARB_CMD 0x00
#define PMIC_ARB_CONFIG 0x04
#define PMIC_ARB_STATUS 0x08
#define PMIC_ARB_WDATA0 0x10
#define PMIC_ARB_WDATA1 0x14
#define PMIC_ARB_RDATA0 0x18
#define PMIC_ARB_RDATA1 0x1C
/* Mapping Table */
#define SPMI_MAPPING_TABLE_REG(N) (0x0B00 + (4 * (N)))
#define SPMI_MAPPING_BIT_INDEX(X) (((X) >> 18) & 0xF)
#define SPMI_MAPPING_BIT_IS_0_FLAG(X) (((X) >> 17) & 0x1)
#define SPMI_MAPPING_BIT_IS_0_RESULT(X) (((X) >> 9) & 0xFF)
#define SPMI_MAPPING_BIT_IS_1_FLAG(X) (((X) >> 8) & 0x1)
#define SPMI_MAPPING_BIT_IS_1_RESULT(X) (((X) >> 0) & 0xFF)
#define SPMI_MAPPING_TABLE_TREE_DEPTH 16 /* Maximum of 16-bits */
#define PMIC_ARB_MAX_PPID BIT(12) /* PPID is 12bit */
#define PMIC_ARB_APID_VALID BIT(15)
#define PMIC_ARB_CHAN_IS_IRQ_OWNER(reg) ((reg) & BIT(24))
#define INVALID_EE 0xFF
/* Ownership Table */
#define SPMI_OWNERSHIP_PERIPH2OWNER(X) ((X) & 0x7)
/* Channel Status fields */
enum pmic_arb_chnl_status {
PMIC_ARB_STATUS_DONE = BIT(0),
PMIC_ARB_STATUS_FAILURE = BIT(1),
PMIC_ARB_STATUS_DENIED = BIT(2),
PMIC_ARB_STATUS_DROPPED = BIT(3),
};
/* Command register fields */
#define PMIC_ARB_CMD_MAX_BYTE_COUNT 8
/* Command Opcodes */
enum pmic_arb_cmd_op_code {
PMIC_ARB_OP_EXT_WRITEL = 0,
PMIC_ARB_OP_EXT_READL = 1,
PMIC_ARB_OP_EXT_WRITE = 2,
PMIC_ARB_OP_RESET = 3,
PMIC_ARB_OP_SLEEP = 4,
PMIC_ARB_OP_SHUTDOWN = 5,
PMIC_ARB_OP_WAKEUP = 6,
PMIC_ARB_OP_AUTHENTICATE = 7,
PMIC_ARB_OP_MSTR_READ = 8,
PMIC_ARB_OP_MSTR_WRITE = 9,
PMIC_ARB_OP_EXT_READ = 13,
PMIC_ARB_OP_WRITE = 14,
PMIC_ARB_OP_READ = 15,
PMIC_ARB_OP_ZERO_WRITE = 16,
};
/*
* PMIC arbiter version 5 uses different register offsets for read/write vs
* observer channels.
*/
enum pmic_arb_channel {
PMIC_ARB_CHANNEL_RW,
PMIC_ARB_CHANNEL_OBS,
};
/* Maximum number of support PMIC peripherals */
#define PMIC_ARB_MAX_PERIPHS 512
#define PMIC_ARB_MAX_PERIPHS_V7 1024
#define PMIC_ARB_TIMEOUT_US 1000
#define PMIC_ARB_MAX_TRANS_BYTES (8)
#define PMIC_ARB_APID_MASK 0xFF
#define PMIC_ARB_PPID_MASK 0xFFF
/* interrupt enable bit */
#define SPMI_PIC_ACC_ENABLE_BIT BIT(0)
#define spec_to_hwirq(slave_id, periph_id, irq_id, apid) \
((((slave_id) & 0xF) << 28) | \
(((periph_id) & 0xFF) << 20) | \
(((irq_id) & 0x7) << 16) | \
(((apid) & 0x3FF) << 0))
#define hwirq_to_sid(hwirq) (((hwirq) >> 28) & 0xF)
#define hwirq_to_per(hwirq) (((hwirq) >> 20) & 0xFF)
#define hwirq_to_irq(hwirq) (((hwirq) >> 16) & 0x7)
#define hwirq_to_apid(hwirq) (((hwirq) >> 0) & 0x3FF)
struct pmic_arb_ver_ops;
struct apid_data {
u16 ppid;
u8 write_ee;
u8 irq_ee;
};
/**
* struct spmi_pmic_arb - SPMI PMIC Arbiter object
*
* @rd_base: on v1 "core", on v2 "observer" register base off DT.
* @wr_base: on v1 "core", on v2 "chnls" register base off DT.
* @intr: address of the SPMI interrupt control registers.
* @cnfg: address of the PMIC Arbiter configuration registers.
* @lock: lock to synchronize accesses.
* @channel: execution environment channel to use for accesses.
* @irq: PMIC ARB interrupt.
* @ee: the current Execution Environment
* @bus_instance: on v7: 0 = primary SPMI bus, 1 = secondary SPMI bus
* @min_apid: minimum APID (used for bounding IRQ search)
* @max_apid: maximum APID
* @base_apid: on v7: minimum APID associated with the particular SPMI
* bus instance
* @apid_count: on v5 and v7: number of APIDs associated with the
* particular SPMI bus instance
* @mapping_table: in-memory copy of PPID -> APID mapping table.
* @domain: irq domain object for PMIC IRQ domain
* @spmic: SPMI controller object
* @ver_ops: version dependent operations.
* @ppid_to_apid: in-memory copy of PPID -> APID mapping table.
* @last_apid: Highest value APID in use
* @apid_data: Table of data for all APIDs
* @max_periphs: Number of elements in apid_data[]
*/
struct spmi_pmic_arb {
void __iomem *rd_base;
void __iomem *wr_base;
void __iomem *intr;
void __iomem *cnfg;
void __iomem *core;
resource_size_t core_size;
raw_spinlock_t lock;
u8 channel;
int irq;
u8 ee;
u32 bus_instance;
u16 min_apid;
u16 max_apid;
u16 base_apid;
int apid_count;
u32 *mapping_table;
DECLARE_BITMAP(mapping_table_valid, PMIC_ARB_MAX_PERIPHS);
struct irq_domain *domain;
struct spmi_controller *spmic;
const struct pmic_arb_ver_ops *ver_ops;
u16 *ppid_to_apid;
u16 last_apid;
struct apid_data *apid_data;
int max_periphs;
};
/**
* struct pmic_arb_ver_ops - version dependent functionality.
*
* @ver_str: version string.
* @ppid_to_apid: finds the apid for a given ppid.
* @non_data_cmd: on v1 issues an spmi non-data command.
* on v2 no HW support, returns -EOPNOTSUPP.
* @offset: on v1 offset of per-ee channel.
* on v2 offset of per-ee and per-ppid channel.
* @fmt_cmd: formats a GENI/SPMI command.
* @owner_acc_status: on v1 address of PMIC_ARB_SPMI_PIC_OWNERm_ACC_STATUSn
* on v2 address of SPMI_PIC_OWNERm_ACC_STATUSn.
* @acc_enable: on v1 address of PMIC_ARB_SPMI_PIC_ACC_ENABLEn
* on v2 address of SPMI_PIC_ACC_ENABLEn.
* @irq_status: on v1 address of PMIC_ARB_SPMI_PIC_IRQ_STATUSn
* on v2 address of SPMI_PIC_IRQ_STATUSn.
* @irq_clear: on v1 address of PMIC_ARB_SPMI_PIC_IRQ_CLEARn
* on v2 address of SPMI_PIC_IRQ_CLEARn.
* @apid_map_offset: offset of PMIC_ARB_REG_CHNLn
* @apid_owner: on v2 and later address of SPMI_PERIPHn_2OWNER_TABLE_REG
*/
struct pmic_arb_ver_ops {
const char *ver_str;
int (*ppid_to_apid)(struct spmi_pmic_arb *pmic_arb, u16 ppid);
/* spmi commands (read_cmd, write_cmd, cmd) functionality */
int (*offset)(struct spmi_pmic_arb *pmic_arb, u8 sid, u16 addr,
enum pmic_arb_channel ch_type);
u32 (*fmt_cmd)(u8 opc, u8 sid, u16 addr, u8 bc);
int (*non_data_cmd)(struct spmi_controller *ctrl, u8 opc, u8 sid);
/* Interrupts controller functionality (offset of PIC registers) */
void __iomem *(*owner_acc_status)(struct spmi_pmic_arb *pmic_arb, u8 m,
u16 n);
void __iomem *(*acc_enable)(struct spmi_pmic_arb *pmic_arb, u16 n);
void __iomem *(*irq_status)(struct spmi_pmic_arb *pmic_arb, u16 n);
void __iomem *(*irq_clear)(struct spmi_pmic_arb *pmic_arb, u16 n);
u32 (*apid_map_offset)(u16 n);
void __iomem *(*apid_owner)(struct spmi_pmic_arb *pmic_arb, u16 n);
};
static inline void pmic_arb_base_write(struct spmi_pmic_arb *pmic_arb,
u32 offset, u32 val)
{
writel_relaxed(val, pmic_arb->wr_base + offset);
}
static inline void pmic_arb_set_rd_cmd(struct spmi_pmic_arb *pmic_arb,
u32 offset, u32 val)
{
writel_relaxed(val, pmic_arb->rd_base + offset);
}
/**
* pmic_arb_read_data: reads pmic-arb's register and copy 1..4 bytes to buf
* @bc: byte count -1. range: 0..3
* @reg: register's address
* @buf: output parameter, length must be bc + 1
*/
static void
pmic_arb_read_data(struct spmi_pmic_arb *pmic_arb, u8 *buf, u32 reg, u8 bc)
{
u32 data = __raw_readl(pmic_arb->rd_base + reg);
memcpy(buf, &data, (bc & 3) + 1);
}
/**
* pmic_arb_write_data: write 1..4 bytes from buf to pmic-arb's register
* @bc: byte-count -1. range: 0..3.
* @reg: register's address.
* @buf: buffer to write. length must be bc + 1.
*/
static void pmic_arb_write_data(struct spmi_pmic_arb *pmic_arb, const u8 *buf,
u32 reg, u8 bc)
{
u32 data = 0;
memcpy(&data, buf, (bc & 3) + 1);
__raw_writel(data, pmic_arb->wr_base + reg);
}
static int pmic_arb_wait_for_done(struct spmi_controller *ctrl,
void __iomem *base, u8 sid, u16 addr,
enum pmic_arb_channel ch_type)
{
struct spmi_pmic_arb *pmic_arb = spmi_controller_get_drvdata(ctrl);
u32 status = 0;
u32 timeout = PMIC_ARB_TIMEOUT_US;
u32 offset;
int rc;
rc = pmic_arb->ver_ops->offset(pmic_arb, sid, addr, ch_type);
if (rc < 0)
return rc;
offset = rc;
offset += PMIC_ARB_STATUS;
while (timeout--) {
status = readl_relaxed(base + offset);
if (status & PMIC_ARB_STATUS_DONE) {
if (status & PMIC_ARB_STATUS_DENIED) {
dev_err(&ctrl->dev, "%s: %#x %#x: transaction denied (%#x)\n",
__func__, sid, addr, status);
return -EPERM;
}
if (status & PMIC_ARB_STATUS_FAILURE) {
dev_err(&ctrl->dev, "%s: %#x %#x: transaction failed (%#x)\n",
__func__, sid, addr, status);
WARN_ON(1);
return -EIO;
}
if (status & PMIC_ARB_STATUS_DROPPED) {
dev_err(&ctrl->dev, "%s: %#x %#x: transaction dropped (%#x)\n",
__func__, sid, addr, status);
return -EIO;
}
return 0;
}
udelay(1);
}
dev_err(&ctrl->dev, "%s: %#x %#x: timeout, status %#x\n",
__func__, sid, addr, status);
return -ETIMEDOUT;
}
static int
pmic_arb_non_data_cmd_v1(struct spmi_controller *ctrl, u8 opc, u8 sid)
{
struct spmi_pmic_arb *pmic_arb = spmi_controller_get_drvdata(ctrl);
unsigned long flags;
u32 cmd;
int rc;
u32 offset;
rc = pmic_arb->ver_ops->offset(pmic_arb, sid, 0, PMIC_ARB_CHANNEL_RW);
if (rc < 0)
return rc;
offset = rc;
cmd = ((opc | 0x40) << 27) | ((sid & 0xf) << 20);
raw_spin_lock_irqsave(&pmic_arb->lock, flags);
pmic_arb_base_write(pmic_arb, offset + PMIC_ARB_CMD, cmd);
rc = pmic_arb_wait_for_done(ctrl, pmic_arb->wr_base, sid, 0,
PMIC_ARB_CHANNEL_RW);
raw_spin_unlock_irqrestore(&pmic_arb->lock, flags);
return rc;
}
static int
pmic_arb_non_data_cmd_v2(struct spmi_controller *ctrl, u8 opc, u8 sid)
{
return -EOPNOTSUPP;
}
/* Non-data command */
static int pmic_arb_cmd(struct spmi_controller *ctrl, u8 opc, u8 sid)
{
struct spmi_pmic_arb *pmic_arb = spmi_controller_get_drvdata(ctrl);
dev_dbg(&ctrl->dev, "cmd op:0x%x sid:%d\n", opc, sid);
/* Check for valid non-data command */
if (opc < SPMI_CMD_RESET || opc > SPMI_CMD_WAKEUP)
return -EINVAL;
return pmic_arb->ver_ops->non_data_cmd(ctrl, opc, sid);
}
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
static int pmic_arb_fmt_read_cmd(struct spmi_pmic_arb *pmic_arb, u8 opc, u8 sid,
u16 addr, size_t len, u32 *cmd, u32 *offset)
{
u8 bc = len - 1;
int rc;
rc = pmic_arb->ver_ops->offset(pmic_arb, sid, addr,
PMIC_ARB_CHANNEL_OBS);
if (rc < 0)
return rc;
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
*offset = rc;
if (bc >= PMIC_ARB_MAX_TRANS_BYTES) {
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
dev_err(&pmic_arb->spmic->dev, "pmic-arb supports 1..%d bytes per trans, but:%zu requested",
PMIC_ARB_MAX_TRANS_BYTES, len);
return -EINVAL;
}
/* Check the opcode */
if (opc >= 0x60 && opc <= 0x7F)
opc = PMIC_ARB_OP_READ;
else if (opc >= 0x20 && opc <= 0x2F)
opc = PMIC_ARB_OP_EXT_READ;
else if (opc >= 0x38 && opc <= 0x3F)
opc = PMIC_ARB_OP_EXT_READL;
else
return -EINVAL;
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
*cmd = pmic_arb->ver_ops->fmt_cmd(opc, sid, addr, bc);
return 0;
}
static int pmic_arb_read_cmd_unlocked(struct spmi_controller *ctrl, u32 cmd,
u32 offset, u8 sid, u16 addr, u8 *buf,
size_t len)
{
struct spmi_pmic_arb *pmic_arb = spmi_controller_get_drvdata(ctrl);
u8 bc = len - 1;
int rc;
pmic_arb_set_rd_cmd(pmic_arb, offset + PMIC_ARB_CMD, cmd);
rc = pmic_arb_wait_for_done(ctrl, pmic_arb->rd_base, sid, addr,
PMIC_ARB_CHANNEL_OBS);
if (rc)
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
return rc;
pmic_arb_read_data(pmic_arb, buf, offset + PMIC_ARB_RDATA0,
min_t(u8, bc, 3));
if (bc > 3)
pmic_arb_read_data(pmic_arb, buf + 4, offset + PMIC_ARB_RDATA1,
bc - 4);
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
return 0;
}
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
static int pmic_arb_read_cmd(struct spmi_controller *ctrl, u8 opc, u8 sid,
u16 addr, u8 *buf, size_t len)
{
struct spmi_pmic_arb *pmic_arb = spmi_controller_get_drvdata(ctrl);
unsigned long flags;
u32 cmd, offset;
int rc;
rc = pmic_arb_fmt_read_cmd(pmic_arb, opc, sid, addr, len, &cmd,
&offset);
if (rc)
return rc;
raw_spin_lock_irqsave(&pmic_arb->lock, flags);
rc = pmic_arb_read_cmd_unlocked(ctrl, cmd, offset, sid, addr, buf, len);
raw_spin_unlock_irqrestore(&pmic_arb->lock, flags);
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
return rc;
}
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
static int pmic_arb_fmt_write_cmd(struct spmi_pmic_arb *pmic_arb, u8 opc,
u8 sid, u16 addr, size_t len, u32 *cmd,
u32 *offset)
{
u8 bc = len - 1;
int rc;
rc = pmic_arb->ver_ops->offset(pmic_arb, sid, addr,
PMIC_ARB_CHANNEL_RW);
if (rc < 0)
return rc;
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
*offset = rc;
if (bc >= PMIC_ARB_MAX_TRANS_BYTES) {
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
dev_err(&pmic_arb->spmic->dev, "pmic-arb supports 1..%d bytes per trans, but:%zu requested",
PMIC_ARB_MAX_TRANS_BYTES, len);
return -EINVAL;
}
/* Check the opcode */
if (opc >= 0x40 && opc <= 0x5F)
opc = PMIC_ARB_OP_WRITE;
else if (opc <= 0x0F)
opc = PMIC_ARB_OP_EXT_WRITE;
else if (opc >= 0x30 && opc <= 0x37)
opc = PMIC_ARB_OP_EXT_WRITEL;
else if (opc >= 0x80)
opc = PMIC_ARB_OP_ZERO_WRITE;
else
return -EINVAL;
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
*cmd = pmic_arb->ver_ops->fmt_cmd(opc, sid, addr, bc);
return 0;
}
static int pmic_arb_write_cmd_unlocked(struct spmi_controller *ctrl, u32 cmd,
u32 offset, u8 sid, u16 addr,
const u8 *buf, size_t len)
{
struct spmi_pmic_arb *pmic_arb = spmi_controller_get_drvdata(ctrl);
u8 bc = len - 1;
/* Write data to FIFOs */
pmic_arb_write_data(pmic_arb, buf, offset + PMIC_ARB_WDATA0,
min_t(u8, bc, 3));
if (bc > 3)
pmic_arb_write_data(pmic_arb, buf + 4, offset + PMIC_ARB_WDATA1,
bc - 4);
/* Start the transaction */
pmic_arb_base_write(pmic_arb, offset + PMIC_ARB_CMD, cmd);
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
return pmic_arb_wait_for_done(ctrl, pmic_arb->wr_base, sid, addr,
PMIC_ARB_CHANNEL_RW);
}
static int pmic_arb_write_cmd(struct spmi_controller *ctrl, u8 opc, u8 sid,
u16 addr, const u8 *buf, size_t len)
{
struct spmi_pmic_arb *pmic_arb = spmi_controller_get_drvdata(ctrl);
unsigned long flags;
u32 cmd, offset;
int rc;
rc = pmic_arb_fmt_write_cmd(pmic_arb, opc, sid, addr, len, &cmd,
&offset);
if (rc)
return rc;
raw_spin_lock_irqsave(&pmic_arb->lock, flags);
rc = pmic_arb_write_cmd_unlocked(ctrl, cmd, offset, sid, addr, buf,
len);
raw_spin_unlock_irqrestore(&pmic_arb->lock, flags);
return rc;
}
static int pmic_arb_masked_write(struct spmi_controller *ctrl, u8 sid, u16 addr,
const u8 *buf, const u8 *mask, size_t len)
{
struct spmi_pmic_arb *pmic_arb = spmi_controller_get_drvdata(ctrl);
u32 read_cmd, read_offset, write_cmd, write_offset;
u8 temp[PMIC_ARB_MAX_TRANS_BYTES];
unsigned long flags;
int rc, i;
rc = pmic_arb_fmt_read_cmd(pmic_arb, SPMI_CMD_EXT_READL, sid, addr, len,
&read_cmd, &read_offset);
if (rc)
return rc;
rc = pmic_arb_fmt_write_cmd(pmic_arb, SPMI_CMD_EXT_WRITEL, sid, addr,
len, &write_cmd, &write_offset);
if (rc)
return rc;
raw_spin_lock_irqsave(&pmic_arb->lock, flags);
rc = pmic_arb_read_cmd_unlocked(ctrl, read_cmd, read_offset, sid, addr,
temp, len);
if (rc)
goto done;
for (i = 0; i < len; i++)
temp[i] = (temp[i] & ~mask[i]) | (buf[i] & mask[i]);
rc = pmic_arb_write_cmd_unlocked(ctrl, write_cmd, write_offset, sid,
addr, temp, len);
done:
raw_spin_unlock_irqrestore(&pmic_arb->lock, flags);
return rc;
}
enum qpnpint_regs {
QPNPINT_REG_RT_STS = 0x10,
QPNPINT_REG_SET_TYPE = 0x11,
QPNPINT_REG_POLARITY_HIGH = 0x12,
QPNPINT_REG_POLARITY_LOW = 0x13,
QPNPINT_REG_LATCHED_CLR = 0x14,
QPNPINT_REG_EN_SET = 0x15,
QPNPINT_REG_EN_CLR = 0x16,
QPNPINT_REG_LATCHED_STS = 0x18,
};
struct spmi_pmic_arb_qpnpint_type {
u8 type; /* 1 -> edge */
u8 polarity_high;
u8 polarity_low;
} __packed;
/* Simplified accessor functions for irqchip callbacks */
static void qpnpint_spmi_write(struct irq_data *d, u8 reg, void *buf,
size_t len)
{
struct spmi_pmic_arb *pmic_arb = irq_data_get_irq_chip_data(d);
u8 sid = hwirq_to_sid(d->hwirq);
u8 per = hwirq_to_per(d->hwirq);
if (pmic_arb_write_cmd(pmic_arb->spmic, SPMI_CMD_EXT_WRITEL, sid,
(per << 8) + reg, buf, len))
dev_err_ratelimited(&pmic_arb->spmic->dev, "failed irqchip transaction on %x\n",
d->irq);
}
static void qpnpint_spmi_read(struct irq_data *d, u8 reg, void *buf, size_t len)
{
struct spmi_pmic_arb *pmic_arb = irq_data_get_irq_chip_data(d);
u8 sid = hwirq_to_sid(d->hwirq);
u8 per = hwirq_to_per(d->hwirq);
if (pmic_arb_read_cmd(pmic_arb->spmic, SPMI_CMD_EXT_READL, sid,
(per << 8) + reg, buf, len))
dev_err_ratelimited(&pmic_arb->spmic->dev, "failed irqchip transaction on %x\n",
d->irq);
}
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
static int qpnpint_spmi_masked_write(struct irq_data *d, u8 reg,
const void *buf, const void *mask,
size_t len)
{
struct spmi_pmic_arb *pmic_arb = irq_data_get_irq_chip_data(d);
u8 sid = hwirq_to_sid(d->hwirq);
u8 per = hwirq_to_per(d->hwirq);
int rc;
rc = pmic_arb_masked_write(pmic_arb->spmic, sid, (per << 8) + reg, buf,
mask, len);
if (rc)
dev_err_ratelimited(&pmic_arb->spmic->dev, "failed irqchip transaction on %x rc=%d\n",
d->irq, rc);
return rc;
}
static void cleanup_irq(struct spmi_pmic_arb *pmic_arb, u16 apid, int id)
{
u16 ppid = pmic_arb->apid_data[apid].ppid;
u8 sid = ppid >> 8;
u8 per = ppid & 0xFF;
u8 irq_mask = BIT(id);
dev_err_ratelimited(&pmic_arb->spmic->dev, "%s apid=%d sid=0x%x per=0x%x irq=%d\n",
__func__, apid, sid, per, id);
writel_relaxed(irq_mask, pmic_arb->ver_ops->irq_clear(pmic_arb, apid));
}
static int periph_interrupt(struct spmi_pmic_arb *pmic_arb, u16 apid)
{
unsigned int irq;
u32 status, id;
int handled = 0;
u8 sid = (pmic_arb->apid_data[apid].ppid >> 8) & 0xF;
u8 per = pmic_arb->apid_data[apid].ppid & 0xFF;
status = readl_relaxed(pmic_arb->ver_ops->irq_status(pmic_arb, apid));
while (status) {
id = ffs(status) - 1;
status &= ~BIT(id);
irq = irq_find_mapping(pmic_arb->domain,
spec_to_hwirq(sid, per, id, apid));
if (irq == 0) {
cleanup_irq(pmic_arb, apid, id);
continue;
}
generic_handle_irq(irq);
handled++;
}
return handled;
}
static void pmic_arb_chained_irq(struct irq_desc *desc)
{
struct spmi_pmic_arb *pmic_arb = irq_desc_get_handler_data(desc);
const struct pmic_arb_ver_ops *ver_ops = pmic_arb->ver_ops;
struct irq_chip *chip = irq_desc_get_chip(desc);
int first = pmic_arb->min_apid;
int last = pmic_arb->max_apid;
/*
* acc_offset will be non-zero for the secondary SPMI bus instance on
* v7 controllers.
*/
int acc_offset = pmic_arb->base_apid >> 5;
u8 ee = pmic_arb->ee;
u32 status, enable, handled = 0;
int i, id, apid;
/* status based dispatch */
bool acc_valid = false;
u32 irq_status = 0;
chained_irq_enter(chip, desc);
for (i = first >> 5; i <= last >> 5; ++i) {
status = readl_relaxed(ver_ops->owner_acc_status(pmic_arb, ee, i - acc_offset));
if (status)
acc_valid = true;
while (status) {
id = ffs(status) - 1;
status &= ~BIT(id);
apid = id + i * 32;
if (apid < first || apid > last) {
WARN_ONCE(true, "spurious spmi irq received for apid=%d\n",
apid);
continue;
}
enable = readl_relaxed(
ver_ops->acc_enable(pmic_arb, apid));
if (enable & SPMI_PIC_ACC_ENABLE_BIT)
if (periph_interrupt(pmic_arb, apid) != 0)
handled++;
}
}
/* ACC_STATUS is empty but IRQ fired check IRQ_STATUS */
if (!acc_valid) {
for (i = first; i <= last; i++) {
/* skip if APPS is not irq owner */
if (pmic_arb->apid_data[i].irq_ee != pmic_arb->ee)
continue;
irq_status = readl_relaxed(
ver_ops->irq_status(pmic_arb, i));
if (irq_status) {
enable = readl_relaxed(
ver_ops->acc_enable(pmic_arb, i));
if (enable & SPMI_PIC_ACC_ENABLE_BIT) {
dev_dbg(&pmic_arb->spmic->dev,
"Dispatching IRQ for apid=%d status=%x\n",
i, irq_status);
if (periph_interrupt(pmic_arb, i) != 0)
handled++;
}
}
}
}
if (handled == 0)
handle_bad_irq(desc);
chained_irq_exit(chip, desc);
}
static void qpnpint_irq_ack(struct irq_data *d)
{
struct spmi_pmic_arb *pmic_arb = irq_data_get_irq_chip_data(d);
u8 irq = hwirq_to_irq(d->hwirq);
u16 apid = hwirq_to_apid(d->hwirq);
u8 data;
writel_relaxed(BIT(irq), pmic_arb->ver_ops->irq_clear(pmic_arb, apid));
data = BIT(irq);
qpnpint_spmi_write(d, QPNPINT_REG_LATCHED_CLR, &data, 1);
}
static void qpnpint_irq_mask(struct irq_data *d)
{
u8 irq = hwirq_to_irq(d->hwirq);
u8 data = BIT(irq);
qpnpint_spmi_write(d, QPNPINT_REG_EN_CLR, &data, 1);
}
static void qpnpint_irq_unmask(struct irq_data *d)
{
struct spmi_pmic_arb *pmic_arb = irq_data_get_irq_chip_data(d);
const struct pmic_arb_ver_ops *ver_ops = pmic_arb->ver_ops;
u8 irq = hwirq_to_irq(d->hwirq);
u16 apid = hwirq_to_apid(d->hwirq);
spmi: pmic-arb: clear the latched status of the interrupt PMIC interrupts each have an internal latched status bit which is not visible from any register. This status bit is set as soon as the conditions specified in the interrupt type and polarity registers are met even if the interrupt is not enabled. When it is set, nothing else changes within the PMIC and no interrupt notification packets are sent. If the internal latched status bit is set when an interrupt is enabled, then the value is immediately propagated into the interrupt latched status register and an interrupt notification packet is sent out from the PMIC over SPMI. This PMIC hardware behavior can lead to a situation where the handler for a level triggered interrupt is called immediately after enable_irq() is called even though the interrupt physically triggered while it was disabled within the genirq framework. This situation takes place if the the interrupt fires twice after calling disable_irq(). The first time it fires, the level flow handler will mask and disregard it. Unfortunately, the second time it fires, the internal latched status bit is set within the PMIC and no further notification is received. When enable_irq() is called later, the interrupt is unmasked (enabled in the PMIC) which results in the PMIC immediately sending an interrupt notification packet out over SPMI. This breaks the semantics of level triggered interrupts within the genirq framework since they should be completely ignored while disabled. The PMIC internal latched status behavior also affects how interrupts are treated during suspend. While entering suspend, all interrupts not specified as wakeup mode are masked. Upon resume, these interrupts are unmasked. Thus if any of the non-wakeup PMIC interrupts fired while the system was suspended, then the PMIC will send interrupt notification packets out via SPMI as soon as they are unmasked during resume. This behavior violates genirq semantics as well since non-wakeup interrupts should be completely ignored during suspend. Modify the qpnpint_irq_unmask() function so that the interrupt latched status clear register is written immediately before the interrupt enable register. This clears the internal latched status bit of the interrupt so that it cannot trigger spuriously immediately upon being enabled. Also, while resuming an irq, an unmask could be called even if it was not previously masked. So, before writing these registers, check if the interrupt is already enabled within the PMIC. If it is, then no further register writes are required. This condition check ensures that a valid latched status register bit is not cleared until it is properly handled. Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org> Signed-off-by: Kiran Gunda <kgunda@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-05-10 14:25:37 +00:00
u8 buf[2];
writel_relaxed(SPMI_PIC_ACC_ENABLE_BIT,
ver_ops->acc_enable(pmic_arb, apid));
spmi: pmic-arb: clear the latched status of the interrupt PMIC interrupts each have an internal latched status bit which is not visible from any register. This status bit is set as soon as the conditions specified in the interrupt type and polarity registers are met even if the interrupt is not enabled. When it is set, nothing else changes within the PMIC and no interrupt notification packets are sent. If the internal latched status bit is set when an interrupt is enabled, then the value is immediately propagated into the interrupt latched status register and an interrupt notification packet is sent out from the PMIC over SPMI. This PMIC hardware behavior can lead to a situation where the handler for a level triggered interrupt is called immediately after enable_irq() is called even though the interrupt physically triggered while it was disabled within the genirq framework. This situation takes place if the the interrupt fires twice after calling disable_irq(). The first time it fires, the level flow handler will mask and disregard it. Unfortunately, the second time it fires, the internal latched status bit is set within the PMIC and no further notification is received. When enable_irq() is called later, the interrupt is unmasked (enabled in the PMIC) which results in the PMIC immediately sending an interrupt notification packet out over SPMI. This breaks the semantics of level triggered interrupts within the genirq framework since they should be completely ignored while disabled. The PMIC internal latched status behavior also affects how interrupts are treated during suspend. While entering suspend, all interrupts not specified as wakeup mode are masked. Upon resume, these interrupts are unmasked. Thus if any of the non-wakeup PMIC interrupts fired while the system was suspended, then the PMIC will send interrupt notification packets out via SPMI as soon as they are unmasked during resume. This behavior violates genirq semantics as well since non-wakeup interrupts should be completely ignored during suspend. Modify the qpnpint_irq_unmask() function so that the interrupt latched status clear register is written immediately before the interrupt enable register. This clears the internal latched status bit of the interrupt so that it cannot trigger spuriously immediately upon being enabled. Also, while resuming an irq, an unmask could be called even if it was not previously masked. So, before writing these registers, check if the interrupt is already enabled within the PMIC. If it is, then no further register writes are required. This condition check ensures that a valid latched status register bit is not cleared until it is properly handled. Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org> Signed-off-by: Kiran Gunda <kgunda@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-05-10 14:25:37 +00:00
qpnpint_spmi_read(d, QPNPINT_REG_EN_SET, &buf[0], 1);
if (!(buf[0] & BIT(irq))) {
/*
* Since the interrupt is currently disabled, write to both the
* LATCHED_CLR and EN_SET registers so that a spurious interrupt
* cannot be triggered when the interrupt is enabled
*/
buf[0] = BIT(irq);
buf[1] = BIT(irq);
qpnpint_spmi_write(d, QPNPINT_REG_LATCHED_CLR, &buf, 2);
}
}
static int qpnpint_irq_set_type(struct irq_data *d, unsigned int flow_type)
{
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
struct spmi_pmic_arb_qpnpint_type type = {0};
struct spmi_pmic_arb_qpnpint_type mask;
irq_flow_handler_t flow_handler;
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
u8 irq_bit = BIT(hwirq_to_irq(d->hwirq));
int rc;
if (flow_type & (IRQF_TRIGGER_RISING | IRQF_TRIGGER_FALLING)) {
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
type.type = irq_bit;
if (flow_type & IRQF_TRIGGER_RISING)
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
type.polarity_high = irq_bit;
if (flow_type & IRQF_TRIGGER_FALLING)
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
type.polarity_low = irq_bit;
flow_handler = handle_edge_irq;
} else {
if ((flow_type & (IRQF_TRIGGER_HIGH)) &&
(flow_type & (IRQF_TRIGGER_LOW)))
return -EINVAL;
if (flow_type & IRQF_TRIGGER_HIGH)
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
type.polarity_high = irq_bit;
else
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
type.polarity_low = irq_bit;
flow_handler = handle_level_irq;
}
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
mask.type = irq_bit;
mask.polarity_high = irq_bit;
mask.polarity_low = irq_bit;
rc = qpnpint_spmi_masked_write(d, QPNPINT_REG_SET_TYPE, &type, &mask,
sizeof(type));
irq_set_handler_locked(d, flow_handler);
spmi: spmi-pmic-arb: fix irq_set_type race condition The qpnpint_irq_set_type() callback function configures the type (edge vs level) and polarity (high, low, or both) of a particular PMIC interrupt within a given peripheral. To do this, it reads the three consecutive IRQ configuration registers, modifies the specified IRQ bit within the register values, and finally writes the three modified register values back to the PMIC. While a spinlock is used to provide mutual exclusion on the SPMI bus during the register read and write calls, there is no locking around the overall read, modify, write sequence. This opens up the possibility of a race condition if two tasks set the type of a PMIC IRQ within the same peripheral simultaneously. When the race condition is encountered, both tasks will read the old value of the registers and IRQ bits set by one of the tasks will be dropped upon the register write of the other task. This then leads to PMIC IRQs being enabled with an incorrect type and polarity configured. Such misconfiguration can lead to an IRQ storm that overwhelms the system and causes it to crash. This race condition and IRQ storm have been observed when using a pair of pm8941-pwrkey devices to handle PMK8350 pwrkey and resin interrupts. The independent devices probe asynchronously in parallel and can simultaneously request and configure PMIC IRQs in the same PMIC peripheral. For a good case, the IRQ configuration calls end up serialized due to timing deltas and the register read/write sequence looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 3. resin probe: SPMI read(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0xC0, 0xC0, 0xC0 The final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0xC0 0x1312: 0xC0 0x1313: 0xC0 0x1314: 0x00 0x1315: 0xC0 For a bad case, the IRQ configuration calls end up occurring simultaneously and the race condition is encountered. The register read/write sequence then looks like this: 1. pwrkey probe: SPMI read(0x1311): 0x00, 0x00, 0x00 2. resin probe: SPMI read(0x1311): 0x00, 0x00, 0x00 3. pwrkey probe: SPMI write(0x1311): 0x80, 0x80, 0x80 4. resin probe: SPMI write(0x1311): 0x40, 0x40, 0x40 In this case, the final register states after both devices have requested and enabled their respective IRQs is thus: 0x1311: 0x40 0x1312: 0x40 0x1313: 0x40 0x1314: 0x00 0x1315: 0xC0 This corresponds to the resin IRQ being configured for both rising and falling edges, as expected. However, the pwrkey IRQ is misconfigured as level type with both polarity high and low set to disabled. The PMIC IRQ triggering hardware treats this particular register configuration as if level low triggering is enabled. The raw pwrkey IRQ signal is low when the power key is not being pressed. Thus, the pwrkey IRQ begins firing continuously in an IRQ storm. Fix the race condition by holding the spmi-pmic-arb spinlock for the duration of the read, modify, write sequence performed in the qpnpint_irq_set_type() function. Split the pmic_arb_read_cmd() and pmic_arb_write_cmd() functions each into three parts so that hardware register IO is decoupled from spinlock locking. This allows a new function pmic_arb_masked_write() to be added which locks the spinlock and then calls register IO functions to perform SPMI read and write commands in a single atomic operation. Link: https://lore.kernel.org/r/20211118034719.28971-1-quic_collinsd@quicinc.com Signed-off-by: David Collins <quic_collinsd@quicinc.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20211216190812.1574801-7-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-12-16 19:08:12 +00:00
return rc;
}
static int qpnpint_irq_set_wake(struct irq_data *d, unsigned int on)
{
struct spmi_pmic_arb *pmic_arb = irq_data_get_irq_chip_data(d);
return irq_set_irq_wake(pmic_arb->irq, on);
}
static int qpnpint_get_irqchip_state(struct irq_data *d,
enum irqchip_irq_state which,
bool *state)
{
u8 irq = hwirq_to_irq(d->hwirq);
u8 status = 0;
if (which != IRQCHIP_STATE_LINE_LEVEL)
return -EINVAL;
qpnpint_spmi_read(d, QPNPINT_REG_RT_STS, &status, 1);
*state = !!(status & BIT(irq));
return 0;
}
static int qpnpint_irq_domain_activate(struct irq_domain *domain,
struct irq_data *d, bool reserve)
{
struct spmi_pmic_arb *pmic_arb = irq_data_get_irq_chip_data(d);
u16 periph = hwirq_to_per(d->hwirq);
u16 apid = hwirq_to_apid(d->hwirq);
u16 sid = hwirq_to_sid(d->hwirq);
u16 irq = hwirq_to_irq(d->hwirq);
spmi: pmic-arb: do not ack and clear peripheral interrupts in cleanup_irq Currently, cleanup_irq() is invoked when a peripheral's interrupt fires and there is no mapping present in the interrupt domain of spmi interrupt controller. The cleanup_irq clears the arbiter bit, clears the pmic interrupt and disables it at the pmic in that order. The last disable in cleanup_irq races with request_irq() in that it stomps over the enable issued by request_irq. Fix this by not writing to the pmic in cleanup_irq. The latched bit will be left set in the pmic, which will not send us more interrupts even if the enable bit stays enabled. When a client wants to request an interrupt, use the activate callback on the irq_domain to clear latched bit. This ensures that the latched, if set due to the above changes in cleanup_irq or when the bootloader leaves it set, gets cleaned up, paving way for upcoming interrupts to trigger. With this, there is a possibility of unwanted triggering of interrupt right after the latched bit is cleared - the interrupt may be left enabled too. To avoid that, clear the enable first followed by clearing the latched bit in the activate callback. Fixes: 6bc546e71e50 ("spmi: pmic-arb: cleanup unrequested irqs") Fixes: 02abec3616c1 ("spmi: pmic-arb: rename pa_xx to pmic_arb_xx and other cleanup") Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org> [collinsd@codeaurora.org: fix merge conflict] Signed-off-by: David Collins <collinsd@codeaurora.org> Signed-off-by: Fenglin Wu <quic_fenglinw@quicinc.com> Link: https://lore.kernel.org/r/1655004286-11493-4-git-send-email-quic_fenglinw@quicinc.com Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20220930005019.2663064-5-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2022-09-30 00:50:13 +00:00
u8 buf;
if (pmic_arb->apid_data[apid].irq_ee != pmic_arb->ee) {
dev_err(&pmic_arb->spmic->dev, "failed to xlate sid = %#x, periph = %#x, irq = %u: ee=%u but owner=%u\n",
sid, periph, irq, pmic_arb->ee,
pmic_arb->apid_data[apid].irq_ee);
return -ENODEV;
}
spmi: pmic-arb: do not ack and clear peripheral interrupts in cleanup_irq Currently, cleanup_irq() is invoked when a peripheral's interrupt fires and there is no mapping present in the interrupt domain of spmi interrupt controller. The cleanup_irq clears the arbiter bit, clears the pmic interrupt and disables it at the pmic in that order. The last disable in cleanup_irq races with request_irq() in that it stomps over the enable issued by request_irq. Fix this by not writing to the pmic in cleanup_irq. The latched bit will be left set in the pmic, which will not send us more interrupts even if the enable bit stays enabled. When a client wants to request an interrupt, use the activate callback on the irq_domain to clear latched bit. This ensures that the latched, if set due to the above changes in cleanup_irq or when the bootloader leaves it set, gets cleaned up, paving way for upcoming interrupts to trigger. With this, there is a possibility of unwanted triggering of interrupt right after the latched bit is cleared - the interrupt may be left enabled too. To avoid that, clear the enable first followed by clearing the latched bit in the activate callback. Fixes: 6bc546e71e50 ("spmi: pmic-arb: cleanup unrequested irqs") Fixes: 02abec3616c1 ("spmi: pmic-arb: rename pa_xx to pmic_arb_xx and other cleanup") Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org> [collinsd@codeaurora.org: fix merge conflict] Signed-off-by: David Collins <collinsd@codeaurora.org> Signed-off-by: Fenglin Wu <quic_fenglinw@quicinc.com> Link: https://lore.kernel.org/r/1655004286-11493-4-git-send-email-quic_fenglinw@quicinc.com Signed-off-by: Stephen Boyd <sboyd@kernel.org> Link: https://lore.kernel.org/r/20220930005019.2663064-5-sboyd@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2022-09-30 00:50:13 +00:00
buf = BIT(irq);
qpnpint_spmi_write(d, QPNPINT_REG_EN_CLR, &buf, 1);
qpnpint_spmi_write(d, QPNPINT_REG_LATCHED_CLR, &buf, 1);
return 0;
}
static struct irq_chip pmic_arb_irqchip = {
.name = "pmic_arb",
.irq_ack = qpnpint_irq_ack,
.irq_mask = qpnpint_irq_mask,
.irq_unmask = qpnpint_irq_unmask,
.irq_set_type = qpnpint_irq_set_type,
.irq_set_wake = qpnpint_irq_set_wake,
.irq_get_irqchip_state = qpnpint_get_irqchip_state,
.flags = IRQCHIP_MASK_ON_SUSPEND,
};
static int qpnpint_irq_domain_translate(struct irq_domain *d,
struct irq_fwspec *fwspec,
unsigned long *out_hwirq,
unsigned int *out_type)
{
struct spmi_pmic_arb *pmic_arb = d->host_data;
u32 *intspec = fwspec->param;
u16 apid, ppid;
int rc;
dev_dbg(&pmic_arb->spmic->dev, "intspec[0] 0x%1x intspec[1] 0x%02x intspec[2] 0x%02x\n",
intspec[0], intspec[1], intspec[2]);
if (irq_domain_get_of_node(d) != pmic_arb->spmic->dev.of_node)
return -EINVAL;
if (fwspec->param_count != 4)
return -EINVAL;
if (intspec[0] > 0xF || intspec[1] > 0xFF || intspec[2] > 0x7)
return -EINVAL;
ppid = intspec[0] << 8 | intspec[1];
rc = pmic_arb->ver_ops->ppid_to_apid(pmic_arb, ppid);
if (rc < 0) {
dev_err(&pmic_arb->spmic->dev, "failed to xlate sid = %#x, periph = %#x, irq = %u rc = %d\n",
intspec[0], intspec[1], intspec[2], rc);
return rc;
}
apid = rc;
/* Keep track of {max,min}_apid for bounding search during interrupt */
if (apid > pmic_arb->max_apid)
pmic_arb->max_apid = apid;
if (apid < pmic_arb->min_apid)
pmic_arb->min_apid = apid;
*out_hwirq = spec_to_hwirq(intspec[0], intspec[1], intspec[2], apid);
*out_type = intspec[3] & IRQ_TYPE_SENSE_MASK;
dev_dbg(&pmic_arb->spmic->dev, "out_hwirq = %lu\n", *out_hwirq);
return 0;
}
spmi: pmic-arb: Set lockdep class for hierarchical irq domains I see the following lockdep splat in the qcom pinctrl driver when attempting to suspend the device. WARNING: possible recursive locking detected 5.4.11 #3 Tainted: G W -------------------------------------------- cat/3074 is trying to acquire lock: ffffff81f49804c0 (&irq_desc_lock_class){-.-.}, at: __irq_get_desc_lock+0x64/0x94 but task is already holding lock: ffffff81f1cc10c0 (&irq_desc_lock_class){-.-.}, at: __irq_get_desc_lock+0x64/0x94 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(&irq_desc_lock_class); lock(&irq_desc_lock_class); *** DEADLOCK *** May be due to missing lock nesting notation 6 locks held by cat/3074: #0: ffffff81f01d9420 (sb_writers#7){.+.+}, at: vfs_write+0xd0/0x1a4 #1: ffffff81bd7d2080 (&of->mutex){+.+.}, at: kernfs_fop_write+0x12c/0x1fc #2: ffffff81f4c322f0 (kn->count#337){.+.+}, at: kernfs_fop_write+0x134/0x1fc #3: ffffffe411a41d60 (system_transition_mutex){+.+.}, at: pm_suspend+0x108/0x348 #4: ffffff81f1c5e970 (&dev->mutex){....}, at: __device_suspend+0x168/0x41c #5: ffffff81f1cc10c0 (&irq_desc_lock_class){-.-.}, at: __irq_get_desc_lock+0x64/0x94 stack backtrace: CPU: 5 PID: 3074 Comm: cat Tainted: G W 5.4.11 #3 Hardware name: Google Cheza (rev3+) (DT) Call trace: dump_backtrace+0x0/0x174 show_stack+0x20/0x2c dump_stack+0xc8/0x124 __lock_acquire+0x460/0x2388 lock_acquire+0x1cc/0x210 _raw_spin_lock_irqsave+0x64/0x80 __irq_get_desc_lock+0x64/0x94 irq_set_irq_wake+0x40/0x144 qpnpint_irq_set_wake+0x28/0x34 set_irq_wake_real+0x40/0x5c irq_set_irq_wake+0x70/0x144 pm8941_pwrkey_suspend+0x34/0x44 platform_pm_suspend+0x34/0x60 dpm_run_callback+0x64/0xcc __device_suspend+0x310/0x41c dpm_suspend+0xf8/0x298 dpm_suspend_start+0x84/0xb4 suspend_devices_and_enter+0xbc/0x620 pm_suspend+0x210/0x348 state_store+0xb0/0x108 kobj_attr_store+0x14/0x24 sysfs_kf_write+0x4c/0x64 kernfs_fop_write+0x15c/0x1fc __vfs_write+0x54/0x18c vfs_write+0xe4/0x1a4 ksys_write+0x7c/0xe4 __arm64_sys_write+0x20/0x2c el0_svc_common+0xa8/0x160 el0_svc_handler+0x7c/0x98 el0_svc+0x8/0xc Set a lockdep class when we map the irq so that irq_set_wake() doesn't warn about a lockdep bug that doesn't exist. Fixes: 12a9eeaebba3 ("spmi: pmic-arb: convert to v2 irq interfaces to support hierarchical IRQ chips") Cc: Douglas Anderson <dianders@chromium.org> Cc: Brian Masney <masneyb@onstation.org> Cc: Lina Iyer <ilina@codeaurora.org> Cc: Maulik Shah <mkshah@codeaurora.org> Cc: Bjorn Andersson <bjorn.andersson@linaro.org> Signed-off-by: Stephen Boyd <swboyd@chromium.org> Link: https://lore.kernel.org/r/20200121183748.68662-1-swboyd@chromium.org Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
2020-01-21 18:37:48 +00:00
static struct lock_class_key qpnpint_irq_lock_class, qpnpint_irq_request_class;
static void qpnpint_irq_domain_map(struct spmi_pmic_arb *pmic_arb,
struct irq_domain *domain, unsigned int virq,
irq_hw_number_t hwirq, unsigned int type)
{
irq_flow_handler_t handler;
dev_dbg(&pmic_arb->spmic->dev, "virq = %u, hwirq = %lu, type = %u\n",
virq, hwirq, type);
if (type & IRQ_TYPE_EDGE_BOTH)
handler = handle_edge_irq;
else
handler = handle_level_irq;
spmi: pmic-arb: Set lockdep class for hierarchical irq domains I see the following lockdep splat in the qcom pinctrl driver when attempting to suspend the device. WARNING: possible recursive locking detected 5.4.11 #3 Tainted: G W -------------------------------------------- cat/3074 is trying to acquire lock: ffffff81f49804c0 (&irq_desc_lock_class){-.-.}, at: __irq_get_desc_lock+0x64/0x94 but task is already holding lock: ffffff81f1cc10c0 (&irq_desc_lock_class){-.-.}, at: __irq_get_desc_lock+0x64/0x94 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(&irq_desc_lock_class); lock(&irq_desc_lock_class); *** DEADLOCK *** May be due to missing lock nesting notation 6 locks held by cat/3074: #0: ffffff81f01d9420 (sb_writers#7){.+.+}, at: vfs_write+0xd0/0x1a4 #1: ffffff81bd7d2080 (&of->mutex){+.+.}, at: kernfs_fop_write+0x12c/0x1fc #2: ffffff81f4c322f0 (kn->count#337){.+.+}, at: kernfs_fop_write+0x134/0x1fc #3: ffffffe411a41d60 (system_transition_mutex){+.+.}, at: pm_suspend+0x108/0x348 #4: ffffff81f1c5e970 (&dev->mutex){....}, at: __device_suspend+0x168/0x41c #5: ffffff81f1cc10c0 (&irq_desc_lock_class){-.-.}, at: __irq_get_desc_lock+0x64/0x94 stack backtrace: CPU: 5 PID: 3074 Comm: cat Tainted: G W 5.4.11 #3 Hardware name: Google Cheza (rev3+) (DT) Call trace: dump_backtrace+0x0/0x174 show_stack+0x20/0x2c dump_stack+0xc8/0x124 __lock_acquire+0x460/0x2388 lock_acquire+0x1cc/0x210 _raw_spin_lock_irqsave+0x64/0x80 __irq_get_desc_lock+0x64/0x94 irq_set_irq_wake+0x40/0x144 qpnpint_irq_set_wake+0x28/0x34 set_irq_wake_real+0x40/0x5c irq_set_irq_wake+0x70/0x144 pm8941_pwrkey_suspend+0x34/0x44 platform_pm_suspend+0x34/0x60 dpm_run_callback+0x64/0xcc __device_suspend+0x310/0x41c dpm_suspend+0xf8/0x298 dpm_suspend_start+0x84/0xb4 suspend_devices_and_enter+0xbc/0x620 pm_suspend+0x210/0x348 state_store+0xb0/0x108 kobj_attr_store+0x14/0x24 sysfs_kf_write+0x4c/0x64 kernfs_fop_write+0x15c/0x1fc __vfs_write+0x54/0x18c vfs_write+0xe4/0x1a4 ksys_write+0x7c/0xe4 __arm64_sys_write+0x20/0x2c el0_svc_common+0xa8/0x160 el0_svc_handler+0x7c/0x98 el0_svc+0x8/0xc Set a lockdep class when we map the irq so that irq_set_wake() doesn't warn about a lockdep bug that doesn't exist. Fixes: 12a9eeaebba3 ("spmi: pmic-arb: convert to v2 irq interfaces to support hierarchical IRQ chips") Cc: Douglas Anderson <dianders@chromium.org> Cc: Brian Masney <masneyb@onstation.org> Cc: Lina Iyer <ilina@codeaurora.org> Cc: Maulik Shah <mkshah@codeaurora.org> Cc: Bjorn Andersson <bjorn.andersson@linaro.org> Signed-off-by: Stephen Boyd <swboyd@chromium.org> Link: https://lore.kernel.org/r/20200121183748.68662-1-swboyd@chromium.org Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
2020-01-21 18:37:48 +00:00
irq_set_lockdep_class(virq, &qpnpint_irq_lock_class,
&qpnpint_irq_request_class);
irq_domain_set_info(domain, virq, hwirq, &pmic_arb_irqchip, pmic_arb,
handler, NULL, NULL);
}
static int qpnpint_irq_domain_alloc(struct irq_domain *domain,
unsigned int virq, unsigned int nr_irqs,
void *data)
{
struct spmi_pmic_arb *pmic_arb = domain->host_data;
struct irq_fwspec *fwspec = data;
irq_hw_number_t hwirq;
unsigned int type;
int ret, i;
ret = qpnpint_irq_domain_translate(domain, fwspec, &hwirq, &type);
if (ret)
return ret;
for (i = 0; i < nr_irqs; i++)
qpnpint_irq_domain_map(pmic_arb, domain, virq + i, hwirq + i,
type);
return 0;
}
static int pmic_arb_ppid_to_apid_v1(struct spmi_pmic_arb *pmic_arb, u16 ppid)
{
u32 *mapping_table = pmic_arb->mapping_table;
int index = 0, i;
u16 apid_valid;
u16 apid;
u32 data;
apid_valid = pmic_arb->ppid_to_apid[ppid];
if (apid_valid & PMIC_ARB_APID_VALID) {
apid = apid_valid & ~PMIC_ARB_APID_VALID;
return apid;
}
for (i = 0; i < SPMI_MAPPING_TABLE_TREE_DEPTH; ++i) {
if (!test_and_set_bit(index, pmic_arb->mapping_table_valid))
mapping_table[index] = readl_relaxed(pmic_arb->cnfg +
SPMI_MAPPING_TABLE_REG(index));
data = mapping_table[index];
if (ppid & BIT(SPMI_MAPPING_BIT_INDEX(data))) {
if (SPMI_MAPPING_BIT_IS_1_FLAG(data)) {
index = SPMI_MAPPING_BIT_IS_1_RESULT(data);
} else {
apid = SPMI_MAPPING_BIT_IS_1_RESULT(data);
pmic_arb->ppid_to_apid[ppid]
= apid | PMIC_ARB_APID_VALID;
pmic_arb->apid_data[apid].ppid = ppid;
return apid;
}
} else {
if (SPMI_MAPPING_BIT_IS_0_FLAG(data)) {
index = SPMI_MAPPING_BIT_IS_0_RESULT(data);
} else {
apid = SPMI_MAPPING_BIT_IS_0_RESULT(data);
pmic_arb->ppid_to_apid[ppid]
= apid | PMIC_ARB_APID_VALID;
pmic_arb->apid_data[apid].ppid = ppid;
return apid;
}
}
}
return -ENODEV;
}
/* v1 offset per ee */
static int pmic_arb_offset_v1(struct spmi_pmic_arb *pmic_arb, u8 sid, u16 addr,
enum pmic_arb_channel ch_type)
{
return 0x800 + 0x80 * pmic_arb->channel;
}
static u16 pmic_arb_find_apid(struct spmi_pmic_arb *pmic_arb, u16 ppid)
{
struct apid_data *apidd = &pmic_arb->apid_data[pmic_arb->last_apid];
u32 regval, offset;
u16 id, apid;
for (apid = pmic_arb->last_apid; ; apid++, apidd++) {
offset = pmic_arb->ver_ops->apid_map_offset(apid);
if (offset >= pmic_arb->core_size)
break;
regval = readl_relaxed(pmic_arb->ver_ops->apid_owner(pmic_arb,
apid));
apidd->irq_ee = SPMI_OWNERSHIP_PERIPH2OWNER(regval);
apidd->write_ee = apidd->irq_ee;
regval = readl_relaxed(pmic_arb->core + offset);
if (!regval)
continue;
id = (regval >> 8) & PMIC_ARB_PPID_MASK;
pmic_arb->ppid_to_apid[id] = apid | PMIC_ARB_APID_VALID;
apidd->ppid = id;
if (id == ppid) {
apid |= PMIC_ARB_APID_VALID;
break;
}
}
pmic_arb->last_apid = apid & ~PMIC_ARB_APID_VALID;
return apid;
}
static int pmic_arb_ppid_to_apid_v2(struct spmi_pmic_arb *pmic_arb, u16 ppid)
{
u16 apid_valid;
apid_valid = pmic_arb->ppid_to_apid[ppid];
if (!(apid_valid & PMIC_ARB_APID_VALID))
apid_valid = pmic_arb_find_apid(pmic_arb, ppid);
if (!(apid_valid & PMIC_ARB_APID_VALID))
return -ENODEV;
return apid_valid & ~PMIC_ARB_APID_VALID;
}
static int pmic_arb_read_apid_map_v5(struct spmi_pmic_arb *pmic_arb)
{
struct apid_data *apidd;
struct apid_data *prev_apidd;
u16 i, apid, ppid, apid_max;
bool valid, is_irq_ee;
u32 regval, offset;
/*
* In order to allow multiple EEs to write to a single PPID in arbiter
* version 5 and 7, there is more than one APID mapped to each PPID.
* The owner field for each of these mappings specifies the EE which is
* allowed to write to the APID. The owner of the last (highest) APID
* which has the IRQ owner bit set for a given PPID will receive
* interrupts from the PPID.
*
* In arbiter version 7, the APID numbering space is divided between
* the primary bus (0) and secondary bus (1) such that:
* APID = 0 to N-1 are assigned to the primary bus
* APID = N to N+M-1 are assigned to the secondary bus
* where N = number of APIDs supported by the primary bus and
* M = number of APIDs supported by the secondary bus
*/
apidd = &pmic_arb->apid_data[pmic_arb->base_apid];
apid_max = pmic_arb->base_apid + pmic_arb->apid_count;
for (i = pmic_arb->base_apid; i < apid_max; i++, apidd++) {
offset = pmic_arb->ver_ops->apid_map_offset(i);
if (offset >= pmic_arb->core_size)
break;
regval = readl_relaxed(pmic_arb->core + offset);
if (!regval)
continue;
ppid = (regval >> 8) & PMIC_ARB_PPID_MASK;
is_irq_ee = PMIC_ARB_CHAN_IS_IRQ_OWNER(regval);
regval = readl_relaxed(pmic_arb->ver_ops->apid_owner(pmic_arb,
i));
apidd->write_ee = SPMI_OWNERSHIP_PERIPH2OWNER(regval);
apidd->irq_ee = is_irq_ee ? apidd->write_ee : INVALID_EE;
valid = pmic_arb->ppid_to_apid[ppid] & PMIC_ARB_APID_VALID;
apid = pmic_arb->ppid_to_apid[ppid] & ~PMIC_ARB_APID_VALID;
prev_apidd = &pmic_arb->apid_data[apid];
if (!valid || apidd->write_ee == pmic_arb->ee) {
/* First PPID mapping or one for this EE */
pmic_arb->ppid_to_apid[ppid] = i | PMIC_ARB_APID_VALID;
} else if (valid && is_irq_ee &&
prev_apidd->write_ee == pmic_arb->ee) {
/*
* Duplicate PPID mapping after the one for this EE;
* override the irq owner
*/
prev_apidd->irq_ee = apidd->irq_ee;
}
apidd->ppid = ppid;
pmic_arb->last_apid = i;
}
/* Dump the mapping table for debug purposes. */
dev_dbg(&pmic_arb->spmic->dev, "PPID APID Write-EE IRQ-EE\n");
for (ppid = 0; ppid < PMIC_ARB_MAX_PPID; ppid++) {
apid = pmic_arb->ppid_to_apid[ppid];
if (apid & PMIC_ARB_APID_VALID) {
apid &= ~PMIC_ARB_APID_VALID;
apidd = &pmic_arb->apid_data[apid];
dev_dbg(&pmic_arb->spmic->dev, "%#03X %3u %2u %2u\n",
ppid, apid, apidd->write_ee, apidd->irq_ee);
}
}
return 0;
}
static int pmic_arb_ppid_to_apid_v5(struct spmi_pmic_arb *pmic_arb, u16 ppid)
{
if (!(pmic_arb->ppid_to_apid[ppid] & PMIC_ARB_APID_VALID))
return -ENODEV;
return pmic_arb->ppid_to_apid[ppid] & ~PMIC_ARB_APID_VALID;
}
/* v2 offset per ppid and per ee */
static int pmic_arb_offset_v2(struct spmi_pmic_arb *pmic_arb, u8 sid, u16 addr,
enum pmic_arb_channel ch_type)
{
u16 apid;
u16 ppid;
int rc;
ppid = sid << 8 | ((addr >> 8) & 0xFF);
rc = pmic_arb_ppid_to_apid_v2(pmic_arb, ppid);
if (rc < 0)
return rc;
apid = rc;
return 0x1000 * pmic_arb->ee + 0x8000 * apid;
}
/*
* v5 offset per ee and per apid for observer channels and per apid for
* read/write channels.
*/
static int pmic_arb_offset_v5(struct spmi_pmic_arb *pmic_arb, u8 sid, u16 addr,
enum pmic_arb_channel ch_type)
{
u16 apid;
int rc;
u32 offset = 0;
u16 ppid = (sid << 8) | (addr >> 8);
rc = pmic_arb_ppid_to_apid_v5(pmic_arb, ppid);
if (rc < 0)
return rc;
apid = rc;
switch (ch_type) {
case PMIC_ARB_CHANNEL_OBS:
offset = 0x10000 * pmic_arb->ee + 0x80 * apid;
break;
case PMIC_ARB_CHANNEL_RW:
if (pmic_arb->apid_data[apid].write_ee != pmic_arb->ee) {
dev_err(&pmic_arb->spmic->dev, "disallowed SPMI write to sid=%u, addr=0x%04X\n",
sid, addr);
return -EPERM;
}
offset = 0x10000 * apid;
break;
}
return offset;
}
/*
* v7 offset per ee and per apid for observer channels and per apid for
* read/write channels.
*/
static int pmic_arb_offset_v7(struct spmi_pmic_arb *pmic_arb, u8 sid, u16 addr,
enum pmic_arb_channel ch_type)
{
u16 apid;
int rc;
u32 offset = 0;
u16 ppid = (sid << 8) | (addr >> 8);
rc = pmic_arb->ver_ops->ppid_to_apid(pmic_arb, ppid);
if (rc < 0)
return rc;
apid = rc;
switch (ch_type) {
case PMIC_ARB_CHANNEL_OBS:
offset = 0x8000 * pmic_arb->ee + 0x20 * apid;
break;
case PMIC_ARB_CHANNEL_RW:
if (pmic_arb->apid_data[apid].write_ee != pmic_arb->ee) {
dev_err(&pmic_arb->spmic->dev, "disallowed SPMI write to sid=%u, addr=0x%04X\n",
sid, addr);
return -EPERM;
}
offset = 0x1000 * apid;
break;
}
return offset;
}
static u32 pmic_arb_fmt_cmd_v1(u8 opc, u8 sid, u16 addr, u8 bc)
{
return (opc << 27) | ((sid & 0xf) << 20) | (addr << 4) | (bc & 0x7);
}
static u32 pmic_arb_fmt_cmd_v2(u8 opc, u8 sid, u16 addr, u8 bc)
{
return (opc << 27) | ((addr & 0xff) << 4) | (bc & 0x7);
}
static void __iomem *
pmic_arb_owner_acc_status_v1(struct spmi_pmic_arb *pmic_arb, u8 m, u16 n)
{
return pmic_arb->intr + 0x20 * m + 0x4 * n;
}
static void __iomem *
pmic_arb_owner_acc_status_v2(struct spmi_pmic_arb *pmic_arb, u8 m, u16 n)
{
return pmic_arb->intr + 0x100000 + 0x1000 * m + 0x4 * n;
}
static void __iomem *
pmic_arb_owner_acc_status_v3(struct spmi_pmic_arb *pmic_arb, u8 m, u16 n)
{
return pmic_arb->intr + 0x200000 + 0x1000 * m + 0x4 * n;
}
static void __iomem *
pmic_arb_owner_acc_status_v5(struct spmi_pmic_arb *pmic_arb, u8 m, u16 n)
{
return pmic_arb->intr + 0x10000 * m + 0x4 * n;
}
static void __iomem *
pmic_arb_owner_acc_status_v7(struct spmi_pmic_arb *pmic_arb, u8 m, u16 n)
{
return pmic_arb->intr + 0x1000 * m + 0x4 * n;
}
static void __iomem *
pmic_arb_acc_enable_v1(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->intr + 0x200 + 0x4 * n;
}
static void __iomem *
pmic_arb_acc_enable_v2(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->intr + 0x1000 * n;
}
static void __iomem *
pmic_arb_acc_enable_v5(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->wr_base + 0x100 + 0x10000 * n;
}
static void __iomem *
pmic_arb_acc_enable_v7(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->wr_base + 0x100 + 0x1000 * n;
}
static void __iomem *
pmic_arb_irq_status_v1(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->intr + 0x600 + 0x4 * n;
}
static void __iomem *
pmic_arb_irq_status_v2(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->intr + 0x4 + 0x1000 * n;
}
static void __iomem *
pmic_arb_irq_status_v5(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->wr_base + 0x104 + 0x10000 * n;
}
static void __iomem *
pmic_arb_irq_status_v7(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->wr_base + 0x104 + 0x1000 * n;
}
static void __iomem *
pmic_arb_irq_clear_v1(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->intr + 0xA00 + 0x4 * n;
}
static void __iomem *
pmic_arb_irq_clear_v2(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->intr + 0x8 + 0x1000 * n;
}
static void __iomem *
pmic_arb_irq_clear_v5(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->wr_base + 0x108 + 0x10000 * n;
}
static void __iomem *
pmic_arb_irq_clear_v7(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->wr_base + 0x108 + 0x1000 * n;
}
static u32 pmic_arb_apid_map_offset_v2(u16 n)
{
return 0x800 + 0x4 * n;
}
static u32 pmic_arb_apid_map_offset_v5(u16 n)
{
return 0x900 + 0x4 * n;
}
static u32 pmic_arb_apid_map_offset_v7(u16 n)
{
return 0x2000 + 0x4 * n;
}
static void __iomem *
pmic_arb_apid_owner_v2(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->cnfg + 0x700 + 0x4 * n;
}
/*
* For arbiter version 7, APID ownership table registers have independent
* numbering space for each SPMI bus instance, so each is indexed starting from
* 0.
*/
static void __iomem *
pmic_arb_apid_owner_v7(struct spmi_pmic_arb *pmic_arb, u16 n)
{
return pmic_arb->cnfg + 0x4 * (n - pmic_arb->base_apid);
}
static const struct pmic_arb_ver_ops pmic_arb_v1 = {
.ver_str = "v1",
.ppid_to_apid = pmic_arb_ppid_to_apid_v1,
.non_data_cmd = pmic_arb_non_data_cmd_v1,
.offset = pmic_arb_offset_v1,
.fmt_cmd = pmic_arb_fmt_cmd_v1,
.owner_acc_status = pmic_arb_owner_acc_status_v1,
.acc_enable = pmic_arb_acc_enable_v1,
.irq_status = pmic_arb_irq_status_v1,
.irq_clear = pmic_arb_irq_clear_v1,
.apid_map_offset = pmic_arb_apid_map_offset_v2,
.apid_owner = pmic_arb_apid_owner_v2,
};
static const struct pmic_arb_ver_ops pmic_arb_v2 = {
.ver_str = "v2",
.ppid_to_apid = pmic_arb_ppid_to_apid_v2,
.non_data_cmd = pmic_arb_non_data_cmd_v2,
.offset = pmic_arb_offset_v2,
.fmt_cmd = pmic_arb_fmt_cmd_v2,
.owner_acc_status = pmic_arb_owner_acc_status_v2,
.acc_enable = pmic_arb_acc_enable_v2,
.irq_status = pmic_arb_irq_status_v2,
.irq_clear = pmic_arb_irq_clear_v2,
.apid_map_offset = pmic_arb_apid_map_offset_v2,
.apid_owner = pmic_arb_apid_owner_v2,
};
static const struct pmic_arb_ver_ops pmic_arb_v3 = {
.ver_str = "v3",
.ppid_to_apid = pmic_arb_ppid_to_apid_v2,
.non_data_cmd = pmic_arb_non_data_cmd_v2,
.offset = pmic_arb_offset_v2,
.fmt_cmd = pmic_arb_fmt_cmd_v2,
.owner_acc_status = pmic_arb_owner_acc_status_v3,
.acc_enable = pmic_arb_acc_enable_v2,
.irq_status = pmic_arb_irq_status_v2,
.irq_clear = pmic_arb_irq_clear_v2,
.apid_map_offset = pmic_arb_apid_map_offset_v2,
.apid_owner = pmic_arb_apid_owner_v2,
};
static const struct pmic_arb_ver_ops pmic_arb_v5 = {
.ver_str = "v5",
.ppid_to_apid = pmic_arb_ppid_to_apid_v5,
.non_data_cmd = pmic_arb_non_data_cmd_v2,
.offset = pmic_arb_offset_v5,
.fmt_cmd = pmic_arb_fmt_cmd_v2,
.owner_acc_status = pmic_arb_owner_acc_status_v5,
.acc_enable = pmic_arb_acc_enable_v5,
.irq_status = pmic_arb_irq_status_v5,
.irq_clear = pmic_arb_irq_clear_v5,
.apid_map_offset = pmic_arb_apid_map_offset_v5,
.apid_owner = pmic_arb_apid_owner_v2,
};
static const struct pmic_arb_ver_ops pmic_arb_v7 = {
.ver_str = "v7",
.ppid_to_apid = pmic_arb_ppid_to_apid_v5,
.non_data_cmd = pmic_arb_non_data_cmd_v2,
.offset = pmic_arb_offset_v7,
.fmt_cmd = pmic_arb_fmt_cmd_v2,
.owner_acc_status = pmic_arb_owner_acc_status_v7,
.acc_enable = pmic_arb_acc_enable_v7,
.irq_status = pmic_arb_irq_status_v7,
.irq_clear = pmic_arb_irq_clear_v7,
.apid_map_offset = pmic_arb_apid_map_offset_v7,
.apid_owner = pmic_arb_apid_owner_v7,
};
static const struct irq_domain_ops pmic_arb_irq_domain_ops = {
.activate = qpnpint_irq_domain_activate,
.alloc = qpnpint_irq_domain_alloc,
.free = irq_domain_free_irqs_common,
.translate = qpnpint_irq_domain_translate,
};
static int spmi_pmic_arb_probe(struct platform_device *pdev)
{
struct spmi_pmic_arb *pmic_arb;
struct spmi_controller *ctrl;
struct resource *res;
void __iomem *core;
u32 *mapping_table;
u32 channel, ee, hw_ver;
int err;
ctrl = devm_spmi_controller_alloc(&pdev->dev, sizeof(*pmic_arb));
if (IS_ERR(ctrl))
return PTR_ERR(ctrl);
pmic_arb = spmi_controller_get_drvdata(ctrl);
pmic_arb->spmic = ctrl;
/*
* Please don't replace this with devm_platform_ioremap_resource() or
* devm_ioremap_resource(). These both result in a call to
* devm_request_mem_region() which prevents multiple mappings of this
* register address range. SoCs with PMIC arbiter v7 may define two
* arbiter devices, for the two physical SPMI interfaces, which share
* some register address ranges (i.e. "core", "obsrvr", and "chnls").
* Ensure that both devices probe successfully by calling devm_ioremap()
* which does not result in a devm_request_mem_region() call.
*/
res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "core");
core = devm_ioremap(&ctrl->dev, res->start, resource_size(res));
if (IS_ERR(core))
return PTR_ERR(core);
pmic_arb->core_size = resource_size(res);
pmic_arb->ppid_to_apid = devm_kcalloc(&ctrl->dev, PMIC_ARB_MAX_PPID,
sizeof(*pmic_arb->ppid_to_apid),
GFP_KERNEL);
if (!pmic_arb->ppid_to_apid)
return -ENOMEM;
hw_ver = readl_relaxed(core + PMIC_ARB_VERSION);
if (hw_ver < PMIC_ARB_VERSION_V2_MIN) {
pmic_arb->ver_ops = &pmic_arb_v1;
pmic_arb->wr_base = core;
pmic_arb->rd_base = core;
} else {
pmic_arb->core = core;
if (hw_ver < PMIC_ARB_VERSION_V3_MIN)
pmic_arb->ver_ops = &pmic_arb_v2;
else if (hw_ver < PMIC_ARB_VERSION_V5_MIN)
pmic_arb->ver_ops = &pmic_arb_v3;
else if (hw_ver < PMIC_ARB_VERSION_V7_MIN)
pmic_arb->ver_ops = &pmic_arb_v5;
else
pmic_arb->ver_ops = &pmic_arb_v7;
res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
"obsrvr");
pmic_arb->rd_base = devm_ioremap(&ctrl->dev, res->start,
resource_size(res));
if (IS_ERR(pmic_arb->rd_base))
return PTR_ERR(pmic_arb->rd_base);
res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
"chnls");
pmic_arb->wr_base = devm_ioremap(&ctrl->dev, res->start,
resource_size(res));
if (IS_ERR(pmic_arb->wr_base))
return PTR_ERR(pmic_arb->wr_base);
}
pmic_arb->max_periphs = PMIC_ARB_MAX_PERIPHS;
if (hw_ver >= PMIC_ARB_VERSION_V7_MIN) {
pmic_arb->max_periphs = PMIC_ARB_MAX_PERIPHS_V7;
/* Optional property for v7: */
of_property_read_u32(pdev->dev.of_node, "qcom,bus-id",
&pmic_arb->bus_instance);
if (pmic_arb->bus_instance > 1) {
dev_err(&pdev->dev, "invalid bus instance (%u) specified\n",
pmic_arb->bus_instance);
return -EINVAL;
}
if (pmic_arb->bus_instance == 0) {
pmic_arb->base_apid = 0;
pmic_arb->apid_count =
readl_relaxed(core + PMIC_ARB_FEATURES) &
PMIC_ARB_FEATURES_PERIPH_MASK;
} else {
pmic_arb->base_apid =
readl_relaxed(core + PMIC_ARB_FEATURES) &
PMIC_ARB_FEATURES_PERIPH_MASK;
pmic_arb->apid_count =
readl_relaxed(core + PMIC_ARB_FEATURES1) &
PMIC_ARB_FEATURES_PERIPH_MASK;
}
if (pmic_arb->base_apid + pmic_arb->apid_count > pmic_arb->max_periphs) {
dev_err(&pdev->dev, "Unsupported APID count %d detected\n",
pmic_arb->base_apid + pmic_arb->apid_count);
return -EINVAL;
}
} else if (hw_ver >= PMIC_ARB_VERSION_V5_MIN) {
pmic_arb->base_apid = 0;
pmic_arb->apid_count = readl_relaxed(core + PMIC_ARB_FEATURES) &
PMIC_ARB_FEATURES_PERIPH_MASK;
if (pmic_arb->apid_count > pmic_arb->max_periphs) {
dev_err(&pdev->dev, "Unsupported APID count %d detected\n",
pmic_arb->apid_count);
return -EINVAL;
}
}
pmic_arb->apid_data = devm_kcalloc(&ctrl->dev, pmic_arb->max_periphs,
sizeof(*pmic_arb->apid_data),
GFP_KERNEL);
if (!pmic_arb->apid_data)
return -ENOMEM;
dev_info(&ctrl->dev, "PMIC arbiter version %s (0x%x)\n",
pmic_arb->ver_ops->ver_str, hw_ver);
res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "intr");
pmic_arb->intr = devm_ioremap_resource(&ctrl->dev, res);
if (IS_ERR(pmic_arb->intr))
return PTR_ERR(pmic_arb->intr);
res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "cnfg");
pmic_arb->cnfg = devm_ioremap_resource(&ctrl->dev, res);
if (IS_ERR(pmic_arb->cnfg))
return PTR_ERR(pmic_arb->cnfg);
pmic_arb->irq = platform_get_irq_byname(pdev, "periph_irq");
if (pmic_arb->irq < 0)
return pmic_arb->irq;
err = of_property_read_u32(pdev->dev.of_node, "qcom,channel", &channel);
if (err) {
dev_err(&pdev->dev, "channel unspecified.\n");
return err;
}
if (channel > 5) {
dev_err(&pdev->dev, "invalid channel (%u) specified.\n",
channel);
return -EINVAL;
}
pmic_arb->channel = channel;
err = of_property_read_u32(pdev->dev.of_node, "qcom,ee", &ee);
if (err) {
dev_err(&pdev->dev, "EE unspecified.\n");
return err;
}
if (ee > 5) {
dev_err(&pdev->dev, "invalid EE (%u) specified\n", ee);
return -EINVAL;
}
pmic_arb->ee = ee;
mapping_table = devm_kcalloc(&ctrl->dev, pmic_arb->max_periphs,
sizeof(*mapping_table), GFP_KERNEL);
if (!mapping_table)
return -ENOMEM;
pmic_arb->mapping_table = mapping_table;
/* Initialize max_apid/min_apid to the opposite bounds, during
* the irq domain translation, we are sure to update these */
pmic_arb->max_apid = 0;
pmic_arb->min_apid = pmic_arb->max_periphs - 1;
platform_set_drvdata(pdev, ctrl);
raw_spin_lock_init(&pmic_arb->lock);
ctrl->cmd = pmic_arb_cmd;
ctrl->read_cmd = pmic_arb_read_cmd;
ctrl->write_cmd = pmic_arb_write_cmd;
if (hw_ver >= PMIC_ARB_VERSION_V5_MIN) {
err = pmic_arb_read_apid_map_v5(pmic_arb);
if (err) {
dev_err(&pdev->dev, "could not read APID->PPID mapping table, rc= %d\n",
err);
return err;
}
}
dev_dbg(&pdev->dev, "adding irq domain\n");
pmic_arb->domain = irq_domain_add_tree(pdev->dev.of_node,
&pmic_arb_irq_domain_ops, pmic_arb);
if (!pmic_arb->domain) {
dev_err(&pdev->dev, "unable to create irq_domain\n");
return -ENOMEM;
}
irq_set_chained_handler_and_data(pmic_arb->irq, pmic_arb_chained_irq,
pmic_arb);
err = spmi_controller_add(ctrl);
if (err)
goto err_domain_remove;
return 0;
err_domain_remove:
irq_set_chained_handler_and_data(pmic_arb->irq, NULL, NULL);
irq_domain_remove(pmic_arb->domain);
return err;
}
static void spmi_pmic_arb_remove(struct platform_device *pdev)
{
struct spmi_controller *ctrl = platform_get_drvdata(pdev);
struct spmi_pmic_arb *pmic_arb = spmi_controller_get_drvdata(ctrl);
spmi_controller_remove(ctrl);
irq_set_chained_handler_and_data(pmic_arb->irq, NULL, NULL);
irq_domain_remove(pmic_arb->domain);
}
static const struct of_device_id spmi_pmic_arb_match_table[] = {
{ .compatible = "qcom,spmi-pmic-arb", },
{},
};
MODULE_DEVICE_TABLE(of, spmi_pmic_arb_match_table);
static struct platform_driver spmi_pmic_arb_driver = {
.probe = spmi_pmic_arb_probe,
.remove_new = spmi_pmic_arb_remove,
.driver = {
.name = "spmi_pmic_arb",
.of_match_table = spmi_pmic_arb_match_table,
},
};
module_platform_driver(spmi_pmic_arb_driver);
MODULE_LICENSE("GPL v2");
MODULE_ALIAS("platform:spmi_pmic_arb");