| Message ID | 20191010182508.22833-2-marex@denx.de |
|---|---|
| State | Changes Requested |
| Delegated to: | David Miller |
| Headers | show |
| Series | [1/2] net: dsa: microchip: Do not reinit mutexes on KSZ87xx | expand |
From: Marek Vasut <marex@denx.de> Date: Thu, 10 Oct 2019 20:25:08 +0200 > The KSZ driver uses one regmap per register width (8/16/32), each with > it's own lock, but accessing the same set of registers. In theory, it > is possible to create a race condition between these regmaps, although > the underlying bus (SPI or I2C) locking should assure nothing bad will > really happen and the accesses would be correct. > > To make the driver do the right thing, add one single shared mutex for > all the regmaps used by the driver instead. This assures that even if > some future hardware is on a bus which does not serialize the accesses > the same way SPI or I2C does, nothing bad will happen. > > Note that the status_mutex was unused and only initied, hence it was > renamed and repurposed as the regmap mutex. > > Signed-off-by: Marek Vasut <marex@denx.de> Applied.
From: David Miller <davem@davemloft.net> Date: Sat, 12 Oct 2019 17:20:55 -0700 (PDT) > From: Marek Vasut <marex@denx.de> > Date: Thu, 10 Oct 2019 20:25:08 +0200 > >> The KSZ driver uses one regmap per register width (8/16/32), each with >> it's own lock, but accessing the same set of registers. In theory, it >> is possible to create a race condition between these regmaps, although >> the underlying bus (SPI or I2C) locking should assure nothing bad will >> really happen and the accesses would be correct. >> >> To make the driver do the right thing, add one single shared mutex for >> all the regmaps used by the driver instead. This assures that even if >> some future hardware is on a bus which does not serialize the accesses >> the same way SPI or I2C does, nothing bad will happen. >> >> Note that the status_mutex was unused and only initied, hence it was >> renamed and repurposed as the regmap mutex. >> >> Signed-off-by: Marek Vasut <marex@denx.de> > > Applied. Actually, both patches reverted. Please test your changes properly: ERROR: "ksz_regmap_unlock" [drivers/net/dsa/microchip/ksz8795_spi.ko] undefined! ERROR: "ksz_regmap_lock" [drivers/net/dsa/microchip/ksz8795_spi.ko] undefined! ERROR: "ksz_regmap_unlock" [drivers/net/dsa/microchip/ksz9477_spi.ko] undefined! ERROR: "ksz_regmap_lock" [drivers/net/dsa/microchip/ksz9477_spi.ko] undefined! ERROR: "ksz_regmap_unlock" [drivers/net/dsa/microchip/ksz9477_i2c.ko] undefined! ERROR: "ksz_regmap_lock" [drivers/net/dsa/microchip/ksz9477_i2c.ko] undefined!
On 10/13/19 2:21 AM, David Miller wrote: > From: David Miller <davem@davemloft.net> > Date: Sat, 12 Oct 2019 17:20:55 -0700 (PDT) > >> From: Marek Vasut <marex@denx.de> >> Date: Thu, 10 Oct 2019 20:25:08 +0200 >> >>> The KSZ driver uses one regmap per register width (8/16/32), each with >>> it's own lock, but accessing the same set of registers. In theory, it >>> is possible to create a race condition between these regmaps, although >>> the underlying bus (SPI or I2C) locking should assure nothing bad will >>> really happen and the accesses would be correct. >>> >>> To make the driver do the right thing, add one single shared mutex for >>> all the regmaps used by the driver instead. This assures that even if >>> some future hardware is on a bus which does not serialize the accesses >>> the same way SPI or I2C does, nothing bad will happen. >>> >>> Note that the status_mutex was unused and only initied, hence it was >>> renamed and repurposed as the regmap mutex. >>> >>> Signed-off-by: Marek Vasut <marex@denx.de> >> >> Applied. > > Actually, both patches reverted. Please test your changes properly: > > ERROR: "ksz_regmap_unlock" [drivers/net/dsa/microchip/ksz8795_spi.ko] undefined! > ERROR: "ksz_regmap_lock" [drivers/net/dsa/microchip/ksz8795_spi.ko] undefined! > ERROR: "ksz_regmap_unlock" [drivers/net/dsa/microchip/ksz9477_spi.ko] undefined! > ERROR: "ksz_regmap_lock" [drivers/net/dsa/microchip/ksz9477_spi.ko] undefined! > ERROR: "ksz_regmap_unlock" [drivers/net/dsa/microchip/ksz9477_i2c.ko] undefined! > ERROR: "ksz_regmap_lock" [drivers/net/dsa/microchip/ksz9477_i2c.ko] undefined! So the test is to compile it as a module ?
From: Marek Vasut <marex@denx.de> Date: Sun, 13 Oct 2019 12:50:15 +0200 > On 10/13/19 2:21 AM, David Miller wrote: >> From: David Miller <davem@davemloft.net> >> Date: Sat, 12 Oct 2019 17:20:55 -0700 (PDT) >> >>> From: Marek Vasut <marex@denx.de> >>> Date: Thu, 10 Oct 2019 20:25:08 +0200 >>> >>>> The KSZ driver uses one regmap per register width (8/16/32), each with >>>> it's own lock, but accessing the same set of registers. In theory, it >>>> is possible to create a race condition between these regmaps, although >>>> the underlying bus (SPI or I2C) locking should assure nothing bad will >>>> really happen and the accesses would be correct. >>>> >>>> To make the driver do the right thing, add one single shared mutex for >>>> all the regmaps used by the driver instead. This assures that even if >>>> some future hardware is on a bus which does not serialize the accesses >>>> the same way SPI or I2C does, nothing bad will happen. >>>> >>>> Note that the status_mutex was unused and only initied, hence it was >>>> renamed and repurposed as the regmap mutex. >>>> >>>> Signed-off-by: Marek Vasut <marex@denx.de> >>> >>> Applied. >> >> Actually, both patches reverted. Please test your changes properly: >> >> ERROR: "ksz_regmap_unlock" [drivers/net/dsa/microchip/ksz8795_spi.ko] undefined! >> ERROR: "ksz_regmap_lock" [drivers/net/dsa/microchip/ksz8795_spi.ko] undefined! >> ERROR: "ksz_regmap_unlock" [drivers/net/dsa/microchip/ksz9477_spi.ko] undefined! >> ERROR: "ksz_regmap_lock" [drivers/net/dsa/microchip/ksz9477_spi.ko] undefined! >> ERROR: "ksz_regmap_unlock" [drivers/net/dsa/microchip/ksz9477_i2c.ko] undefined! >> ERROR: "ksz_regmap_lock" [drivers/net/dsa/microchip/ksz9477_i2c.ko] undefined! > > So the test is to compile it as a module ? The test is to compile it in all relevant possible configurations. As a module, statically, and with dependent modules both static and modular as is relevant.
diff --git a/drivers/net/dsa/microchip/ksz8795_spi.c b/drivers/net/dsa/microchip/ksz8795_spi.c index d0f8153e86b7..614404c40cba 100644 --- a/drivers/net/dsa/microchip/ksz8795_spi.c +++ b/drivers/net/dsa/microchip/ksz8795_spi.c @@ -26,6 +26,7 @@ KSZ_REGMAP_TABLE(ksz8795, 16, SPI_ADDR_SHIFT, static int ksz8795_spi_probe(struct spi_device *spi) { struct ksz_device *dev; + struct regmap_config rc; int i, ret; dev = ksz_switch_alloc(&spi->dev, spi); @@ -33,9 +34,9 @@ static int ksz8795_spi_probe(struct spi_device *spi) return -ENOMEM; for (i = 0; i < ARRAY_SIZE(ksz8795_regmap_config); i++) { - dev->regmap[i] = devm_regmap_init_spi(spi, - &ksz8795_regmap_config - [i]); + rc = ksz8795_regmap_config[i]; + rc.lock_arg = &dev->regmap_mutex; + dev->regmap[i] = devm_regmap_init_spi(spi, &rc); if (IS_ERR(dev->regmap[i])) { ret = PTR_ERR(dev->regmap[i]); dev_err(&spi->dev, diff --git a/drivers/net/dsa/microchip/ksz9477_i2c.c b/drivers/net/dsa/microchip/ksz9477_i2c.c index b0a1595d780d..2cb178675b83 100644 --- a/drivers/net/dsa/microchip/ksz9477_i2c.c +++ b/drivers/net/dsa/microchip/ksz9477_i2c.c @@ -18,6 +18,7 @@ static int ksz9477_i2c_probe(struct i2c_client *i2c, const struct i2c_device_id *i2c_id) { struct ksz_device *dev; + struct regmap_config rc; int i, ret; dev = ksz_switch_alloc(&i2c->dev, i2c); @@ -25,8 +26,9 @@ static int ksz9477_i2c_probe(struct i2c_client *i2c, return -ENOMEM; for (i = 0; i < ARRAY_SIZE(ksz9477_regmap_config); i++) { - dev->regmap[i] = devm_regmap_init_i2c(i2c, - &ksz9477_regmap_config[i]); + rc = ksz9477_regmap_config[i]; + rc.lock_arg = &dev->regmap_mutex; + dev->regmap[i] = devm_regmap_init_i2c(i2c, &rc); if (IS_ERR(dev->regmap[i])) { ret = PTR_ERR(dev->regmap[i]); dev_err(&i2c->dev, diff --git a/drivers/net/dsa/microchip/ksz9477_spi.c b/drivers/net/dsa/microchip/ksz9477_spi.c index f4198d6f72be..a39d2373eb3a 100644 --- a/drivers/net/dsa/microchip/ksz9477_spi.c +++ b/drivers/net/dsa/microchip/ksz9477_spi.c @@ -25,6 +25,7 @@ KSZ_REGMAP_TABLE(ksz9477, 32, SPI_ADDR_SHIFT, static int ksz9477_spi_probe(struct spi_device *spi) { struct ksz_device *dev; + struct regmap_config rc; int i, ret; dev = ksz_switch_alloc(&spi->dev, spi); @@ -32,8 +33,9 @@ static int ksz9477_spi_probe(struct spi_device *spi) return -ENOMEM; for (i = 0; i < ARRAY_SIZE(ksz9477_regmap_config); i++) { - dev->regmap[i] = devm_regmap_init_spi(spi, - &ksz9477_regmap_config[i]); + rc = ksz9477_regmap_config[i]; + rc.lock_arg = &dev->regmap_mutex; + dev->regmap[i] = devm_regmap_init_spi(spi, &rc); if (IS_ERR(dev->regmap[i])) { ret = PTR_ERR(dev->regmap[i]); dev_err(&spi->dev, diff --git a/drivers/net/dsa/microchip/ksz_common.c b/drivers/net/dsa/microchip/ksz_common.c index b0b870f0c252..21e483f6e786 100644 --- a/drivers/net/dsa/microchip/ksz_common.c +++ b/drivers/net/dsa/microchip/ksz_common.c @@ -416,6 +416,18 @@ struct ksz_device *ksz_switch_alloc(struct device *base, void *priv) } EXPORT_SYMBOL(ksz_switch_alloc); +void ksz_regmap_lock(void *__mtx) +{ + struct mutex *mtx = __mtx; + mutex_lock(mtx); +} + +void ksz_regmap_unlock(void *__mtx) +{ + struct mutex *mtx = __mtx; + mutex_unlock(mtx); +} + int ksz_switch_register(struct ksz_device *dev, const struct ksz_dev_ops *ops) { @@ -436,7 +448,7 @@ int ksz_switch_register(struct ksz_device *dev, } mutex_init(&dev->dev_mutex); - mutex_init(&dev->stats_mutex); + mutex_init(&dev->regmap_mutex); mutex_init(&dev->alu_mutex); mutex_init(&dev->vlan_mutex); diff --git a/drivers/net/dsa/microchip/ksz_common.h b/drivers/net/dsa/microchip/ksz_common.h index dd60d0837fc6..c21410353ea8 100644 --- a/drivers/net/dsa/microchip/ksz_common.h +++ b/drivers/net/dsa/microchip/ksz_common.h @@ -47,7 +47,7 @@ struct ksz_device { const char *name; struct mutex dev_mutex; /* device access */ - struct mutex stats_mutex; /* status access */ + struct mutex regmap_mutex; /* regmap access */ struct mutex alu_mutex; /* ALU access */ struct mutex vlan_mutex; /* vlan access */ const struct ksz_dev_ops *dev_ops; @@ -145,6 +145,9 @@ struct ksz_dev_ops { void (*exit)(struct ksz_device *dev); }; +void ksz_regmap_lock(void *__mtx); +void ksz_regmap_unlock(void *__mtx); + struct ksz_device *ksz_switch_alloc(struct device *base, void *priv); int ksz_switch_register(struct ksz_device *dev, const struct ksz_dev_ops *ops); @@ -314,6 +317,8 @@ static inline void ksz_pwrite32(struct ksz_device *dev, int port, int offset, .write_flag_mask = \ KSZ_SPI_OP_FLAG_MASK(KSZ_SPI_OP_WR, swp, \ regbits, regpad), \ + .lock = ksz_regmap_lock, \ + .unlock = ksz_regmap_unlock, \ .reg_format_endian = REGMAP_ENDIAN_BIG, \ .val_format_endian = REGMAP_ENDIAN_BIG \ }
The KSZ driver uses one regmap per register width (8/16/32), each with it's own lock, but accessing the same set of registers. In theory, it is possible to create a race condition between these regmaps, although the underlying bus (SPI or I2C) locking should assure nothing bad will really happen and the accesses would be correct. To make the driver do the right thing, add one single shared mutex for all the regmaps used by the driver instead. This assures that even if some future hardware is on a bus which does not serialize the accesses the same way SPI or I2C does, nothing bad will happen. Note that the status_mutex was unused and only initied, hence it was renamed and repurposed as the regmap mutex. Signed-off-by: Marek Vasut <marex@denx.de> Cc: Andrew Lunn <andrew@lunn.ch> Cc: David S. Miller <davem@davemloft.net> Cc: Florian Fainelli <f.fainelli@gmail.com> Cc: George McCollister <george.mccollister@gmail.com> Cc: Tristram Ha <Tristram.Ha@microchip.com> Cc: Woojung Huh <woojung.huh@microchip.com> --- drivers/net/dsa/microchip/ksz8795_spi.c | 7 ++++--- drivers/net/dsa/microchip/ksz9477_i2c.c | 6 ++++-- drivers/net/dsa/microchip/ksz9477_spi.c | 6 ++++-- drivers/net/dsa/microchip/ksz_common.c | 14 +++++++++++++- drivers/net/dsa/microchip/ksz_common.h | 7 ++++++- 5 files changed, 31 insertions(+), 9 deletions(-)