| Message ID | 20200821024006.3399663-1-furquan@google.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | f492ffe414a7cddc4fd7351563300bc8711ca187 |
| Headers | show |
| Series | [v2] drivers: input: Use single i2c_transfer transaction when using RM_CMD_BANK_SWITCH | expand |
Hi Furquan, On Thu, Aug 20, 2020 at 07:40:06PM -0700, Furquan Shaikh wrote: > On an AMD chromebook, where the same I2C bus is shared by both Raydium > touchscreen and a trackpad device, it is observed that interleaving of > I2C messages when `raydium_i2c_read_message()` is called leads to the > Raydium touch IC reporting incorrect information. This is the sequence > that was observed to result in the above issue: > > * I2C write to Raydium device for RM_CMD_BANK_SWITCH > * I2C write to trackpad device > * I2C read from trackpad device > * I2C write to Raydium device for setting address > * I2C read from Raydium device >>>> This provides incorrect > information > > This change adds a new helper function `raydium_i2c_xfer()` that > performs I2C transactions to the Raydium device. It uses the register > address to decide if RM_CMD_BANK_SWITCH header needs to be sent to the > device (i.e. if register address is greater than 255, then bank switch > header is sent before the rest of the transaction). Additionally, it > ensures that all the I2C operations performed as part of > `raydium_i2c_xfer()` are done as a single i2c_transfer. This > guarantees that no other transactions are initiated to any other > device on the same bus in between. Additionally, > `raydium_i2c_{send|read}*` functions are refactored to use this new > helper function. > > Verified with the patch across multiple reboots (>100) that the > information reported by the Raydium touchscreen device during probe > is correct. > > Signed-off-by: Furquan Shaikh <furquan@google.com> > > --- > v2: Added a new helper function raydium_i2c_xfer so that all read and > send transfers are handled using the same path. This helper function > uses register address > 255 to decide whether to send > RM_CMD_BANK_SWITCH command. Additionally, __packed keyword is moved > to be placed after the structure defintion. > > drivers/input/touchscreen/raydium_i2c_ts.c | 132 +++++++++------------ > 1 file changed, 58 insertions(+), 74 deletions(-) > > diff --git a/drivers/input/touchscreen/raydium_i2c_ts.c b/drivers/input/touchscreen/raydium_i2c_ts.c > index fe245439adee..261e4a290836 100644 > --- a/drivers/input/touchscreen/raydium_i2c_ts.c > +++ b/drivers/input/touchscreen/raydium_i2c_ts.c > @@ -51,6 +51,7 @@ > > /* Touch relative info */ > #define RM_MAX_RETRIES 3 > +#define RM_RETRY_DELAY_MS 20 > #define RM_MAX_TOUCH_NUM 10 > #define RM_BOOT_DELAY_MS 100 > > @@ -136,114 +137,98 @@ struct raydium_data { > bool wake_irq_enabled; > }; > > -static int raydium_i2c_send(struct i2c_client *client, > - u8 addr, const void *data, size_t len) > +static int raydium_i2c_xfer(struct i2c_client *client, u32 addr, void *data, > + size_t len, bool is_read) > { > - u8 *buf; > - int tries = 0; > - int ret; > - > - buf = kmalloc(len + 1, GFP_KERNEL); > - if (!buf) > - return -ENOMEM; > - > - buf[0] = addr; > - memcpy(buf + 1, data, len); > - > - do { > - ret = i2c_master_send(client, buf, len + 1); > - if (likely(ret == len + 1)) > - break; > - > - msleep(20); > - } while (++tries < RM_MAX_RETRIES); > - > - kfree(buf); > - > - if (unlikely(ret != len + 1)) { > - if (ret >= 0) > - ret = -EIO; > - dev_err(&client->dev, "%s failed: %d\n", __func__, ret); > - return ret; > - } > + struct raydium_bank_switch_header { > + u8 cmd; > + __be32 be_addr; > + } __packed header = { > + .cmd = RM_CMD_BANK_SWITCH, > + .be_addr = cpu_to_be32(addr), > + }; > > - return 0; > -} > + u8 reg_addr = addr & 0xff; > > -static int raydium_i2c_read(struct i2c_client *client, > - u8 addr, void *data, size_t len) > -{ > struct i2c_msg xfer[] = { > + { > + .addr = client->addr, > + .len = sizeof(header), > + .buf = (u8 *)&header, > + }, > { > .addr = client->addr, > .len = 1, > - .buf = &addr, > + .buf = ®_addr, > }, > { > .addr = client->addr, > - .flags = I2C_M_RD, > .len = len, > .buf = data, > + .flags = is_read ? I2C_M_RD : 0, > } > }; > + > + /* > + * If address is greater than 255, then RM_CMD_BANK_SWITCH needs to be > + * sent first. Else, skip the header i.e. xfer[0]. > + */ > + int xfer_start_idx = (addr > 0xff) ? 0 : 1; > + size_t xfer_count = ARRAY_SIZE(xfer) - xfer_start_idx; > int ret; > > - ret = i2c_transfer(client->adapter, xfer, ARRAY_SIZE(xfer)); > - if (unlikely(ret != ARRAY_SIZE(xfer))) > - return ret < 0 ? ret : -EIO; > + ret = i2c_transfer(client->adapter, &xfer[xfer_start_idx], xfer_count); > + if (likely(ret == xfer_count)) > + return 0; > > - return 0; > + return -EIO; We are losing real error here, I'll change this to return ret < 0 ? ret : -EIO; > } > > -static int raydium_i2c_read_message(struct i2c_client *client, > - u32 addr, void *data, size_t len) > +static int raydium_i2c_send(struct i2c_client *client, u32 addr, > + const void *data, size_t len) > { > - __be32 be_addr; > + int tries = 0; > + int ret; > + > + do { > + ret = raydium_i2c_xfer(client, addr, (void *)data, len, false); > + if (likely(ret == 0)) > + return 0; > + > + msleep(RM_RETRY_DELAY_MS); > + } while (++tries < RM_MAX_RETRIES); > + > + dev_err(&client->dev, "%s failed\n", __func__); > + return -EIO; Again losing real error here, I'll rename ret to error and do return error; here. > +} > + > +static int raydium_i2c_read(struct i2c_client *client, u32 addr, void *data, > + size_t len) > +{ > + int ret; > size_t xfer_len; > - int error; > > while (len) { > xfer_len = min_t(size_t, len, RM_MAX_READ_SIZE); > + ret = raydium_i2c_xfer(client, addr, data, len, true); I think this needs to be xfer_len, not len. I can fix it up locally if you agree. Thanks.
Hi Dmitry, On Tue, Sep 8, 2020 at 5:44 PM Dmitry Torokhov <dmitry.torokhov@gmail.com> wrote: > > Hi Furquan, > > On Thu, Aug 20, 2020 at 07:40:06PM -0700, Furquan Shaikh wrote: > > On an AMD chromebook, where the same I2C bus is shared by both Raydium > > touchscreen and a trackpad device, it is observed that interleaving of > > I2C messages when `raydium_i2c_read_message()` is called leads to the > > Raydium touch IC reporting incorrect information. This is the sequence > > that was observed to result in the above issue: > > > > * I2C write to Raydium device for RM_CMD_BANK_SWITCH > > * I2C write to trackpad device > > * I2C read from trackpad device > > * I2C write to Raydium device for setting address > > * I2C read from Raydium device >>>> This provides incorrect > > information > > > > This change adds a new helper function `raydium_i2c_xfer()` that > > performs I2C transactions to the Raydium device. It uses the register > > address to decide if RM_CMD_BANK_SWITCH header needs to be sent to the > > device (i.e. if register address is greater than 255, then bank switch > > header is sent before the rest of the transaction). Additionally, it > > ensures that all the I2C operations performed as part of > > `raydium_i2c_xfer()` are done as a single i2c_transfer. This > > guarantees that no other transactions are initiated to any other > > device on the same bus in between. Additionally, > > `raydium_i2c_{send|read}*` functions are refactored to use this new > > helper function. > > > > Verified with the patch across multiple reboots (>100) that the > > information reported by the Raydium touchscreen device during probe > > is correct. > > > > Signed-off-by: Furquan Shaikh <furquan@google.com> > > > > --- > > v2: Added a new helper function raydium_i2c_xfer so that all read and > > send transfers are handled using the same path. This helper function > > uses register address > 255 to decide whether to send > > RM_CMD_BANK_SWITCH command. Additionally, __packed keyword is moved > > to be placed after the structure defintion. > > > > drivers/input/touchscreen/raydium_i2c_ts.c | 132 +++++++++------------ > > 1 file changed, 58 insertions(+), 74 deletions(-) > > > > diff --git a/drivers/input/touchscreen/raydium_i2c_ts.c b/drivers/input/touchscreen/raydium_i2c_ts.c > > index fe245439adee..261e4a290836 100644 > > --- a/drivers/input/touchscreen/raydium_i2c_ts.c > > +++ b/drivers/input/touchscreen/raydium_i2c_ts.c > > @@ -51,6 +51,7 @@ > > > > /* Touch relative info */ > > #define RM_MAX_RETRIES 3 > > +#define RM_RETRY_DELAY_MS 20 > > #define RM_MAX_TOUCH_NUM 10 > > #define RM_BOOT_DELAY_MS 100 > > > > @@ -136,114 +137,98 @@ struct raydium_data { > > bool wake_irq_enabled; > > }; > > > > -static int raydium_i2c_send(struct i2c_client *client, > > - u8 addr, const void *data, size_t len) > > +static int raydium_i2c_xfer(struct i2c_client *client, u32 addr, void *data, > > + size_t len, bool is_read) > > { > > - u8 *buf; > > - int tries = 0; > > - int ret; > > - > > - buf = kmalloc(len + 1, GFP_KERNEL); > > - if (!buf) > > - return -ENOMEM; > > - > > - buf[0] = addr; > > - memcpy(buf + 1, data, len); > > - > > - do { > > - ret = i2c_master_send(client, buf, len + 1); > > - if (likely(ret == len + 1)) > > - break; > > - > > - msleep(20); > > - } while (++tries < RM_MAX_RETRIES); > > - > > - kfree(buf); > > - > > - if (unlikely(ret != len + 1)) { > > - if (ret >= 0) > > - ret = -EIO; > > - dev_err(&client->dev, "%s failed: %d\n", __func__, ret); > > - return ret; > > - } > > + struct raydium_bank_switch_header { > > + u8 cmd; > > + __be32 be_addr; > > + } __packed header = { > > + .cmd = RM_CMD_BANK_SWITCH, > > + .be_addr = cpu_to_be32(addr), > > + }; > > > > - return 0; > > -} > > + u8 reg_addr = addr & 0xff; > > > > -static int raydium_i2c_read(struct i2c_client *client, > > - u8 addr, void *data, size_t len) > > -{ > > struct i2c_msg xfer[] = { > > + { > > + .addr = client->addr, > > + .len = sizeof(header), > > + .buf = (u8 *)&header, > > + }, > > { > > .addr = client->addr, > > .len = 1, > > - .buf = &addr, > > + .buf = ®_addr, > > }, > > { > > .addr = client->addr, > > - .flags = I2C_M_RD, > > .len = len, > > .buf = data, > > + .flags = is_read ? I2C_M_RD : 0, > > } > > }; > > + > > + /* > > + * If address is greater than 255, then RM_CMD_BANK_SWITCH needs to be > > + * sent first. Else, skip the header i.e. xfer[0]. > > + */ > > + int xfer_start_idx = (addr > 0xff) ? 0 : 1; > > + size_t xfer_count = ARRAY_SIZE(xfer) - xfer_start_idx; > > int ret; > > > > - ret = i2c_transfer(client->adapter, xfer, ARRAY_SIZE(xfer)); > > - if (unlikely(ret != ARRAY_SIZE(xfer))) > > - return ret < 0 ? ret : -EIO; > > + ret = i2c_transfer(client->adapter, &xfer[xfer_start_idx], xfer_count); > > + if (likely(ret == xfer_count)) > > + return 0; > > > > - return 0; > > + return -EIO; > > We are losing real error here, I'll change this to > > return ret < 0 ? ret : -EIO; > > > } Looks good. > > > > -static int raydium_i2c_read_message(struct i2c_client *client, > > - u32 addr, void *data, size_t len) > > +static int raydium_i2c_send(struct i2c_client *client, u32 addr, > > + const void *data, size_t len) > > { > > - __be32 be_addr; > > + int tries = 0; > > + int ret; > > + > > + do { > > + ret = raydium_i2c_xfer(client, addr, (void *)data, len, false); > > + if (likely(ret == 0)) > > + return 0; > > + > > + msleep(RM_RETRY_DELAY_MS); > > + } while (++tries < RM_MAX_RETRIES); > > + > > + dev_err(&client->dev, "%s failed\n", __func__); > > + return -EIO; > > Again losing real error here, I'll rename ret to error and do > > return error; > > here. Looks good. > > > +} > > + > > +static int raydium_i2c_read(struct i2c_client *client, u32 addr, void *data, > > + size_t len) > > +{ > > + int ret; > > size_t xfer_len; > > - int error; > > > > while (len) { > > xfer_len = min_t(size_t, len, RM_MAX_READ_SIZE); > > + ret = raydium_i2c_xfer(client, addr, data, len, true); > > I think this needs to be xfer_len, not len. Good catch. Yes, this needs to be xfer_len. > > I can fix it up locally if you agree. The suggestions look good to me. Thank you! - Furquan > > Thanks. > > -- > Dmitry
On Sun, Sep 13, 2020 at 11:04:48PM -0700, Furquan Shaikh wrote: > Hi Dmitry, > > On Tue, Sep 8, 2020 at 5:44 PM Dmitry Torokhov > <dmitry.torokhov@gmail.com> wrote: > > > > Hi Furquan, > > > > On Thu, Aug 20, 2020 at 07:40:06PM -0700, Furquan Shaikh wrote: > > > On an AMD chromebook, where the same I2C bus is shared by both Raydium > > > touchscreen and a trackpad device, it is observed that interleaving of > > > I2C messages when `raydium_i2c_read_message()` is called leads to the > > > Raydium touch IC reporting incorrect information. This is the sequence > > > that was observed to result in the above issue: > > > > > > * I2C write to Raydium device for RM_CMD_BANK_SWITCH > > > * I2C write to trackpad device > > > * I2C read from trackpad device > > > * I2C write to Raydium device for setting address > > > * I2C read from Raydium device >>>> This provides incorrect > > > information > > > > > > This change adds a new helper function `raydium_i2c_xfer()` that > > > performs I2C transactions to the Raydium device. It uses the register > > > address to decide if RM_CMD_BANK_SWITCH header needs to be sent to the > > > device (i.e. if register address is greater than 255, then bank switch > > > header is sent before the rest of the transaction). Additionally, it > > > ensures that all the I2C operations performed as part of > > > `raydium_i2c_xfer()` are done as a single i2c_transfer. This > > > guarantees that no other transactions are initiated to any other > > > device on the same bus in between. Additionally, > > > `raydium_i2c_{send|read}*` functions are refactored to use this new > > > helper function. > > > > > > Verified with the patch across multiple reboots (>100) that the > > > information reported by the Raydium touchscreen device during probe > > > is correct. > > > > > > Signed-off-by: Furquan Shaikh <furquan@google.com> > > > > > > --- > > > v2: Added a new helper function raydium_i2c_xfer so that all read and > > > send transfers are handled using the same path. This helper function > > > uses register address > 255 to decide whether to send > > > RM_CMD_BANK_SWITCH command. Additionally, __packed keyword is moved > > > to be placed after the structure defintion. > > > > > > drivers/input/touchscreen/raydium_i2c_ts.c | 132 +++++++++------------ > > > 1 file changed, 58 insertions(+), 74 deletions(-) > > > > > > diff --git a/drivers/input/touchscreen/raydium_i2c_ts.c b/drivers/input/touchscreen/raydium_i2c_ts.c > > > index fe245439adee..261e4a290836 100644 > > > --- a/drivers/input/touchscreen/raydium_i2c_ts.c > > > +++ b/drivers/input/touchscreen/raydium_i2c_ts.c > > > @@ -51,6 +51,7 @@ > > > > > > /* Touch relative info */ > > > #define RM_MAX_RETRIES 3 > > > +#define RM_RETRY_DELAY_MS 20 > > > #define RM_MAX_TOUCH_NUM 10 > > > #define RM_BOOT_DELAY_MS 100 > > > > > > @@ -136,114 +137,98 @@ struct raydium_data { > > > bool wake_irq_enabled; > > > }; > > > > > > -static int raydium_i2c_send(struct i2c_client *client, > > > - u8 addr, const void *data, size_t len) > > > +static int raydium_i2c_xfer(struct i2c_client *client, u32 addr, void *data, > > > + size_t len, bool is_read) > > > { > > > - u8 *buf; > > > - int tries = 0; > > > - int ret; > > > - > > > - buf = kmalloc(len + 1, GFP_KERNEL); > > > - if (!buf) > > > - return -ENOMEM; > > > - > > > - buf[0] = addr; > > > - memcpy(buf + 1, data, len); > > > - > > > - do { > > > - ret = i2c_master_send(client, buf, len + 1); > > > - if (likely(ret == len + 1)) > > > - break; > > > - > > > - msleep(20); > > > - } while (++tries < RM_MAX_RETRIES); > > > - > > > - kfree(buf); > > > - > > > - if (unlikely(ret != len + 1)) { > > > - if (ret >= 0) > > > - ret = -EIO; > > > - dev_err(&client->dev, "%s failed: %d\n", __func__, ret); > > > - return ret; > > > - } > > > + struct raydium_bank_switch_header { > > > + u8 cmd; > > > + __be32 be_addr; > > > + } __packed header = { > > > + .cmd = RM_CMD_BANK_SWITCH, > > > + .be_addr = cpu_to_be32(addr), > > > + }; > > > > > > - return 0; > > > -} > > > + u8 reg_addr = addr & 0xff; > > > > > > -static int raydium_i2c_read(struct i2c_client *client, > > > - u8 addr, void *data, size_t len) > > > -{ > > > struct i2c_msg xfer[] = { > > > + { > > > + .addr = client->addr, > > > + .len = sizeof(header), > > > + .buf = (u8 *)&header, > > > + }, > > > { > > > .addr = client->addr, > > > .len = 1, > > > - .buf = &addr, > > > + .buf = ®_addr, > > > }, > > > { > > > .addr = client->addr, > > > - .flags = I2C_M_RD, > > > .len = len, > > > .buf = data, > > > + .flags = is_read ? I2C_M_RD : 0, > > > } > > > }; > > > + > > > + /* > > > + * If address is greater than 255, then RM_CMD_BANK_SWITCH needs to be > > > + * sent first. Else, skip the header i.e. xfer[0]. > > > + */ > > > + int xfer_start_idx = (addr > 0xff) ? 0 : 1; > > > + size_t xfer_count = ARRAY_SIZE(xfer) - xfer_start_idx; > > > int ret; > > > > > > - ret = i2c_transfer(client->adapter, xfer, ARRAY_SIZE(xfer)); > > > - if (unlikely(ret != ARRAY_SIZE(xfer))) > > > - return ret < 0 ? ret : -EIO; > > > + ret = i2c_transfer(client->adapter, &xfer[xfer_start_idx], xfer_count); > > > + if (likely(ret == xfer_count)) > > > + return 0; > > > > > > - return 0; > > > + return -EIO; > > > > We are losing real error here, I'll change this to > > > > return ret < 0 ? ret : -EIO; > > > > > } > > Looks good. > > > > > > > -static int raydium_i2c_read_message(struct i2c_client *client, > > > - u32 addr, void *data, size_t len) > > > +static int raydium_i2c_send(struct i2c_client *client, u32 addr, > > > + const void *data, size_t len) > > > { > > > - __be32 be_addr; > > > + int tries = 0; > > > + int ret; > > > + > > > + do { > > > + ret = raydium_i2c_xfer(client, addr, (void *)data, len, false); > > > + if (likely(ret == 0)) > > > + return 0; > > > + > > > + msleep(RM_RETRY_DELAY_MS); > > > + } while (++tries < RM_MAX_RETRIES); > > > + > > > + dev_err(&client->dev, "%s failed\n", __func__); > > > + return -EIO; > > > > Again losing real error here, I'll rename ret to error and do > > > > return error; > > > > here. > > Looks good. > > > > > > +} > > > + > > > +static int raydium_i2c_read(struct i2c_client *client, u32 addr, void *data, > > > + size_t len) > > > +{ > > > + int ret; > > > size_t xfer_len; > > > - int error; > > > > > > while (len) { > > > xfer_len = min_t(size_t, len, RM_MAX_READ_SIZE); > > > + ret = raydium_i2c_xfer(client, addr, data, len, true); > > > > I think this needs to be xfer_len, not len. > > Good catch. Yes, this needs to be xfer_len. > > > > > I can fix it up locally if you agree. > > The suggestions look good to me. Thank you! Great, applied.
diff --git a/drivers/input/touchscreen/raydium_i2c_ts.c b/drivers/input/touchscreen/raydium_i2c_ts.c index fe245439adee..261e4a290836 100644 --- a/drivers/input/touchscreen/raydium_i2c_ts.c +++ b/drivers/input/touchscreen/raydium_i2c_ts.c @@ -51,6 +51,7 @@ /* Touch relative info */ #define RM_MAX_RETRIES 3 +#define RM_RETRY_DELAY_MS 20 #define RM_MAX_TOUCH_NUM 10 #define RM_BOOT_DELAY_MS 100 @@ -136,114 +137,98 @@ struct raydium_data { bool wake_irq_enabled; }; -static int raydium_i2c_send(struct i2c_client *client, - u8 addr, const void *data, size_t len) +static int raydium_i2c_xfer(struct i2c_client *client, u32 addr, void *data, + size_t len, bool is_read) { - u8 *buf; - int tries = 0; - int ret; - - buf = kmalloc(len + 1, GFP_KERNEL); - if (!buf) - return -ENOMEM; - - buf[0] = addr; - memcpy(buf + 1, data, len); - - do { - ret = i2c_master_send(client, buf, len + 1); - if (likely(ret == len + 1)) - break; - - msleep(20); - } while (++tries < RM_MAX_RETRIES); - - kfree(buf); - - if (unlikely(ret != len + 1)) { - if (ret >= 0) - ret = -EIO; - dev_err(&client->dev, "%s failed: %d\n", __func__, ret); - return ret; - } + struct raydium_bank_switch_header { + u8 cmd; + __be32 be_addr; + } __packed header = { + .cmd = RM_CMD_BANK_SWITCH, + .be_addr = cpu_to_be32(addr), + }; - return 0; -} + u8 reg_addr = addr & 0xff; -static int raydium_i2c_read(struct i2c_client *client, - u8 addr, void *data, size_t len) -{ struct i2c_msg xfer[] = { + { + .addr = client->addr, + .len = sizeof(header), + .buf = (u8 *)&header, + }, { .addr = client->addr, .len = 1, - .buf = &addr, + .buf = ®_addr, }, { .addr = client->addr, - .flags = I2C_M_RD, .len = len, .buf = data, + .flags = is_read ? I2C_M_RD : 0, } }; + + /* + * If address is greater than 255, then RM_CMD_BANK_SWITCH needs to be + * sent first. Else, skip the header i.e. xfer[0]. + */ + int xfer_start_idx = (addr > 0xff) ? 0 : 1; + size_t xfer_count = ARRAY_SIZE(xfer) - xfer_start_idx; int ret; - ret = i2c_transfer(client->adapter, xfer, ARRAY_SIZE(xfer)); - if (unlikely(ret != ARRAY_SIZE(xfer))) - return ret < 0 ? ret : -EIO; + ret = i2c_transfer(client->adapter, &xfer[xfer_start_idx], xfer_count); + if (likely(ret == xfer_count)) + return 0; - return 0; + return -EIO; } -static int raydium_i2c_read_message(struct i2c_client *client, - u32 addr, void *data, size_t len) +static int raydium_i2c_send(struct i2c_client *client, u32 addr, + const void *data, size_t len) { - __be32 be_addr; + int tries = 0; + int ret; + + do { + ret = raydium_i2c_xfer(client, addr, (void *)data, len, false); + if (likely(ret == 0)) + return 0; + + msleep(RM_RETRY_DELAY_MS); + } while (++tries < RM_MAX_RETRIES); + + dev_err(&client->dev, "%s failed\n", __func__); + return -EIO; +} + +static int raydium_i2c_read(struct i2c_client *client, u32 addr, void *data, + size_t len) +{ + int ret; size_t xfer_len; - int error; while (len) { xfer_len = min_t(size_t, len, RM_MAX_READ_SIZE); + ret = raydium_i2c_xfer(client, addr, data, len, true); - be_addr = cpu_to_be32(addr); - - error = raydium_i2c_send(client, RM_CMD_BANK_SWITCH, - &be_addr, sizeof(be_addr)); - if (!error) - error = raydium_i2c_read(client, addr & 0xff, - data, xfer_len); - if (error) - return error; + if (unlikely(ret != 0)) + return ret; len -= xfer_len; data += xfer_len; addr += xfer_len; } - return 0; } -static int raydium_i2c_send_message(struct i2c_client *client, - u32 addr, const void *data, size_t len) -{ - __be32 be_addr = cpu_to_be32(addr); - int error; - - error = raydium_i2c_send(client, RM_CMD_BANK_SWITCH, - &be_addr, sizeof(be_addr)); - if (!error) - error = raydium_i2c_send(client, addr & 0xff, data, len); - - return error; -} - static int raydium_i2c_sw_reset(struct i2c_client *client) { const u8 soft_rst_cmd = 0x01; int error; - error = raydium_i2c_send_message(client, RM_RESET_MSG_ADDR, - &soft_rst_cmd, sizeof(soft_rst_cmd)); + error = raydium_i2c_send(client, RM_RESET_MSG_ADDR, &soft_rst_cmd, + sizeof(soft_rst_cmd)); if (error) { dev_err(&client->dev, "software reset failed: %d\n", error); return error; @@ -295,9 +280,8 @@ static int raydium_i2c_query_ts_info(struct raydium_data *ts) if (error) continue; - error = raydium_i2c_read_message(client, - le32_to_cpu(query_bank_addr), - &ts->info, sizeof(ts->info)); + error = raydium_i2c_read(client, le32_to_cpu(query_bank_addr), + &ts->info, sizeof(ts->info)); if (error) continue; @@ -834,7 +818,7 @@ static irqreturn_t raydium_i2c_irq(int irq, void *_dev) if (ts->boot_mode != RAYDIUM_TS_MAIN) goto out; - error = raydium_i2c_read_message(ts->client, ts->data_bank_addr, + error = raydium_i2c_read(ts->client, ts->data_bank_addr, ts->report_data, ts->pkg_size); if (error) goto out;
On an AMD chromebook, where the same I2C bus is shared by both Raydium touchscreen and a trackpad device, it is observed that interleaving of I2C messages when `raydium_i2c_read_message()` is called leads to the Raydium touch IC reporting incorrect information. This is the sequence that was observed to result in the above issue: * I2C write to Raydium device for RM_CMD_BANK_SWITCH * I2C write to trackpad device * I2C read from trackpad device * I2C write to Raydium device for setting address * I2C read from Raydium device >>>> This provides incorrect information This change adds a new helper function `raydium_i2c_xfer()` that performs I2C transactions to the Raydium device. It uses the register address to decide if RM_CMD_BANK_SWITCH header needs to be sent to the device (i.e. if register address is greater than 255, then bank switch header is sent before the rest of the transaction). Additionally, it ensures that all the I2C operations performed as part of `raydium_i2c_xfer()` are done as a single i2c_transfer. This guarantees that no other transactions are initiated to any other device on the same bus in between. Additionally, `raydium_i2c_{send|read}*` functions are refactored to use this new helper function. Verified with the patch across multiple reboots (>100) that the information reported by the Raydium touchscreen device during probe is correct. Signed-off-by: Furquan Shaikh <furquan@google.com> --- v2: Added a new helper function raydium_i2c_xfer so that all read and send transfers are handled using the same path. This helper function uses register address > 255 to decide whether to send RM_CMD_BANK_SWITCH command. Additionally, __packed keyword is moved to be placed after the structure defintion. drivers/input/touchscreen/raydium_i2c_ts.c | 132 +++++++++------------ 1 file changed, 58 insertions(+), 74 deletions(-)