| Message ID | 1446545174-14193-2-git-send-email-vigneshr@ti.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 3 November 2015 at 11:06, Vignesh R <vigneshr@ti.com> wrote: > In addition to providing direct access to SPI bus, some spi controller > hardwares (like ti-qspi) provide special memory mapped port > to accesses SPI flash devices in order to increase read performance. > This means the controller can automatically send the SPI signals > required to read data from the SPI flash device. > For this, spi controller needs to know flash specific information like > read command to use, dummy bytes and address width. Once these settings > are populated in hardware registers, any read accesses to flash's memory > map region(SoC specific) through memcpy (or mem-to mem DMA copy) will be > handled by controller hardware. The hardware will automatically generate > SPI signals required to read data from flash and present it to CPU/DMA. > > Introduce spi_mtd_mmap_read() interface to support memory mapped read > over SPI flash devices. SPI master drivers can implement this callback to > support memory mapped read interfaces. m25p80 flash driver and other > flash drivers can call this to request memory mapped read. The interface > should only be used MTD flashes and cannot be used with other SPI devices. > > Signed-off-by: Vignesh R <vigneshr@ti.com> > --- > drivers/spi/spi.c | 35 +++++++++++++++++++++++++++++++++++ > include/linux/spi/spi.h | 23 +++++++++++++++++++++++ > 2 files changed, 58 insertions(+) > > diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c > index a5f53de813d3..5a5c7a7d47f2 100644 > --- a/drivers/spi/spi.c > +++ b/drivers/spi/spi.c > @@ -1059,6 +1059,7 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) > } > } > > + mutex_lock(&master->mmap_lock_mutex); > trace_spi_message_start(master->cur_msg); > > if (master->prepare_message) { > @@ -1068,6 +1069,7 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) > "failed to prepare message: %d\n", ret); > master->cur_msg->status = ret; > spi_finalize_current_message(master); > + mutex_unlock(&master->mmap_lock_mutex); > return; > } > master->cur_msg_prepared = true; > @@ -1077,6 +1079,7 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) > if (ret) { > master->cur_msg->status = ret; > spi_finalize_current_message(master); > + mutex_unlock(&master->mmap_lock_mutex); > return; > } > > @@ -1084,8 +1087,10 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) > if (ret) { > dev_err(&master->dev, > "failed to transfer one message from queue\n"); > + mutex_unlock(&master->mmap_lock_mutex); > return; > } > + mutex_unlock(&master->mmap_lock_mutex); > } > > /** > @@ -1732,6 +1737,7 @@ int spi_register_master(struct spi_master *master) > spin_lock_init(&master->queue_lock); > spin_lock_init(&master->bus_lock_spinlock); > mutex_init(&master->bus_lock_mutex); > + mutex_init(&master->mmap_lock_mutex); > master->bus_lock_flag = 0; > init_completion(&master->xfer_completion); > if (!master->max_dma_len) > @@ -2237,6 +2243,35 @@ int spi_async_locked(struct spi_device *spi, struct spi_message *message) > EXPORT_SYMBOL_GPL(spi_async_locked); > > > +int spi_mtd_mmap_read(struct spi_device *spi, loff_t from, size_t len, > + size_t *retlen, u_char *buf, u8 read_opcode, > + u8 addr_width, u8 dummy_bytes) > + > +{ > + struct spi_master *master = spi->master; > + int ret; > + > + if (master->auto_runtime_pm) { > + ret = pm_runtime_get_sync(master->dev.parent); > + if (ret < 0) { > + dev_err(&master->dev, "Failed to power device: %d\n", > + ret); > + goto err; > + } > + } > + mutex_lock(&master->mmap_lock_mutex); > + ret = master->spi_mtd_mmap_read(spi, from, len, retlen, buf, > + read_opcode, addr_width, > + dummy_bytes); > + mutex_unlock(&master->mmap_lock_mutex); > + if (master->auto_runtime_pm) > + pm_runtime_put(master->dev.parent); > + > +err: > + return ret; > +} > +EXPORT_SYMBOL_GPL(spi_mtd_mmap_read); > + > /*-------------------------------------------------------------------------*/ > > /* Utility methods for SPI master protocol drivers, layered on > diff --git a/include/linux/spi/spi.h b/include/linux/spi/spi.h > index 6b00f18f5e6b..0a6d8ad57357 100644 > --- a/include/linux/spi/spi.h > +++ b/include/linux/spi/spi.h > @@ -297,6 +297,7 @@ static inline void spi_unregister_driver(struct spi_driver *sdrv) > * @flags: other constraints relevant to this driver > * @bus_lock_spinlock: spinlock for SPI bus locking > * @bus_lock_mutex: mutex for SPI bus locking > + * @mmap_lock_mutex: mutex for locking SPI bus when mmap transfer is on. Any reason to not use the bus_lock_mutex here? The bus is busy as much during mmap transfer as it is during any other transfer. Thanks Michal -- To unsubscribe from this list: send the line "unsubscribe linux-spi" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi, On 11/03/2015 04:49 PM, Michal Suchanek wrote: > On 3 November 2015 at 11:06, Vignesh R <vigneshr@ti.com> wrote: >> In addition to providing direct access to SPI bus, some spi controller >> hardwares (like ti-qspi) provide special memory mapped port >> to accesses SPI flash devices in order to increase read performance. >> This means the controller can automatically send the SPI signals >> required to read data from the SPI flash device. >> For this, spi controller needs to know flash specific information like >> read command to use, dummy bytes and address width. Once these settings >> are populated in hardware registers, any read accesses to flash's memory >> map region(SoC specific) through memcpy (or mem-to mem DMA copy) will be >> handled by controller hardware. The hardware will automatically generate >> SPI signals required to read data from flash and present it to CPU/DMA. >> >> Introduce spi_mtd_mmap_read() interface to support memory mapped read >> over SPI flash devices. SPI master drivers can implement this callback to >> support memory mapped read interfaces. m25p80 flash driver and other >> flash drivers can call this to request memory mapped read. The interface >> should only be used MTD flashes and cannot be used with other SPI devices. >> >> Signed-off-by: Vignesh R <vigneshr@ti.com> >> --- >> drivers/spi/spi.c | 35 +++++++++++++++++++++++++++++++++++ >> include/linux/spi/spi.h | 23 +++++++++++++++++++++++ >> 2 files changed, 58 insertions(+) >> >> diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c >> index a5f53de813d3..5a5c7a7d47f2 100644 >> --- a/drivers/spi/spi.c >> +++ b/drivers/spi/spi.c >> @@ -1059,6 +1059,7 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) >> } >> } >> >> + mutex_lock(&master->mmap_lock_mutex); >> trace_spi_message_start(master->cur_msg); >> >> if (master->prepare_message) { >> @@ -1068,6 +1069,7 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) >> "failed to prepare message: %d\n", ret); >> master->cur_msg->status = ret; >> spi_finalize_current_message(master); >> + mutex_unlock(&master->mmap_lock_mutex); >> return; >> } >> master->cur_msg_prepared = true; >> @@ -1077,6 +1079,7 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) >> if (ret) { >> master->cur_msg->status = ret; >> spi_finalize_current_message(master); >> + mutex_unlock(&master->mmap_lock_mutex); >> return; >> } >> >> @@ -1084,8 +1087,10 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) >> if (ret) { >> dev_err(&master->dev, >> "failed to transfer one message from queue\n"); >> + mutex_unlock(&master->mmap_lock_mutex); >> return; >> } >> + mutex_unlock(&master->mmap_lock_mutex); >> } >> >> /** >> @@ -1732,6 +1737,7 @@ int spi_register_master(struct spi_master *master) >> spin_lock_init(&master->queue_lock); >> spin_lock_init(&master->bus_lock_spinlock); >> mutex_init(&master->bus_lock_mutex); >> + mutex_init(&master->mmap_lock_mutex); >> master->bus_lock_flag = 0; >> init_completion(&master->xfer_completion); >> if (!master->max_dma_len) >> @@ -2237,6 +2243,35 @@ int spi_async_locked(struct spi_device *spi, struct spi_message *message) >> EXPORT_SYMBOL_GPL(spi_async_locked); >> >> >> +int spi_mtd_mmap_read(struct spi_device *spi, loff_t from, size_t len, >> + size_t *retlen, u_char *buf, u8 read_opcode, >> + u8 addr_width, u8 dummy_bytes) >> + >> +{ >> + struct spi_master *master = spi->master; >> + int ret; >> + >> + if (master->auto_runtime_pm) { >> + ret = pm_runtime_get_sync(master->dev.parent); >> + if (ret < 0) { >> + dev_err(&master->dev, "Failed to power device: %d\n", >> + ret); >> + goto err; >> + } >> + } >> + mutex_lock(&master->mmap_lock_mutex); >> + ret = master->spi_mtd_mmap_read(spi, from, len, retlen, buf, >> + read_opcode, addr_width, >> + dummy_bytes); >> + mutex_unlock(&master->mmap_lock_mutex); >> + if (master->auto_runtime_pm) >> + pm_runtime_put(master->dev.parent); >> + >> +err: >> + return ret; >> +} >> +EXPORT_SYMBOL_GPL(spi_mtd_mmap_read); >> + >> /*-------------------------------------------------------------------------*/ >> >> /* Utility methods for SPI master protocol drivers, layered on >> diff --git a/include/linux/spi/spi.h b/include/linux/spi/spi.h >> index 6b00f18f5e6b..0a6d8ad57357 100644 >> --- a/include/linux/spi/spi.h >> +++ b/include/linux/spi/spi.h >> @@ -297,6 +297,7 @@ static inline void spi_unregister_driver(struct spi_driver *sdrv) >> * @flags: other constraints relevant to this driver >> * @bus_lock_spinlock: spinlock for SPI bus locking >> * @bus_lock_mutex: mutex for SPI bus locking >> + * @mmap_lock_mutex: mutex for locking SPI bus when mmap transfer is on. > > Any reason to not use the bus_lock_mutex here? The bus is busy as much > during mmap transfer as it is during any other transfer. If bus_lock_mutex is used, it will unnecessarily block queueing of new messages when __spi_pump_messages() is calling transfer_one_message() even in the case where there is no mmap mode. Hence, I chose to add new mutex. But looks like it doesn't matter much as __spi_sync() anyways blocks till message is transferred. I will drop the new mutex and use bus_lock_mutex in v3. Thanks!
On Tue, Nov 03, 2015 at 03:36:10PM +0530, Vignesh R wrote: > + } > + mutex_lock(&master->mmap_lock_mutex); > + ret = master->spi_mtd_mmap_read(spi, from, len, retlen, buf, > + read_opcode, addr_width, > + dummy_bytes); > + mutex_unlock(&master->mmap_lock_mutex); > + if (master->auto_runtime_pm) > + pm_runtime_put(master->dev.parent); It's a bit worrying that this doesn't sync with the message queue except via the mutex: this means that we might be out of order with respect to any asynchronous transfers that are happening on the device. I'm not sure that this is a practical problem, though there is some risk of unfair scheduling that would have to be under extreme load and it might make sense to prioritise reads anyway.
On 11/04/2015 08:09 PM, Mark Brown wrote: > On Tue, Nov 03, 2015 at 03:36:10PM +0530, Vignesh R wrote: > >> + } >> + mutex_lock(&master->mmap_lock_mutex); >> + ret = master->spi_mtd_mmap_read(spi, from, len, retlen, buf, >> + read_opcode, addr_width, >> + dummy_bytes); >> + mutex_unlock(&master->mmap_lock_mutex); >> + if (master->auto_runtime_pm) >> + pm_runtime_put(master->dev.parent); > > It's a bit worrying that this doesn't sync with the message queue except > via the mutex: this means that we might be out of order with respect to > any asynchronous transfers that are happening on the device. I'm not > sure that this is a practical problem, though there is some risk of > unfair scheduling that would have to be under extreme load and it might > make sense to prioritise reads anyway. > Since mmap interface is used only by mtd flash drivers and since almost all mtd flash devices use synchronous transfers (spi_sync()), IMO, there wont be out of order problem wrt mtd flashes. But mmap read might delay transfers queued for non mtd flash devices. It is difficult to wait for all transfers already queued to be pumped out by __spi_pump_messages() and then do the mmap transfer. Do you have any thoughts on how to sync with message queue?
On Thu, Nov 05, 2015 at 05:59:36PM +0530, Vignesh R wrote: > On 11/04/2015 08:09 PM, Mark Brown wrote: > > It's a bit worrying that this doesn't sync with the message queue except > > via the mutex: this means that we might be out of order with respect to > > any asynchronous transfers that are happening on the device. I'm not > > sure that this is a practical problem, though there is some risk of > > unfair scheduling that would have to be under extreme load and it might > > make sense to prioritise reads anyway. > Since mmap interface is used only by mtd flash drivers and since almost > all mtd flash devices use synchronous transfers (spi_sync()), IMO, there > wont be out of order problem wrt mtd flashes. > But mmap read might delay transfers queued for non mtd flash devices. It > is difficult to wait for all transfers already queued to be pumped out > by __spi_pump_messages() and then do the mmap transfer. Do you have any > thoughts on how to sync with message queue? Like I say I think it probably doesn't matter. The main issue would be with other devices sharing the bus with a flash chip.
diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c index a5f53de813d3..5a5c7a7d47f2 100644 --- a/drivers/spi/spi.c +++ b/drivers/spi/spi.c @@ -1059,6 +1059,7 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) } } + mutex_lock(&master->mmap_lock_mutex); trace_spi_message_start(master->cur_msg); if (master->prepare_message) { @@ -1068,6 +1069,7 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) "failed to prepare message: %d\n", ret); master->cur_msg->status = ret; spi_finalize_current_message(master); + mutex_unlock(&master->mmap_lock_mutex); return; } master->cur_msg_prepared = true; @@ -1077,6 +1079,7 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) if (ret) { master->cur_msg->status = ret; spi_finalize_current_message(master); + mutex_unlock(&master->mmap_lock_mutex); return; } @@ -1084,8 +1087,10 @@ static void __spi_pump_messages(struct spi_master *master, bool in_kthread) if (ret) { dev_err(&master->dev, "failed to transfer one message from queue\n"); + mutex_unlock(&master->mmap_lock_mutex); return; } + mutex_unlock(&master->mmap_lock_mutex); } /** @@ -1732,6 +1737,7 @@ int spi_register_master(struct spi_master *master) spin_lock_init(&master->queue_lock); spin_lock_init(&master->bus_lock_spinlock); mutex_init(&master->bus_lock_mutex); + mutex_init(&master->mmap_lock_mutex); master->bus_lock_flag = 0; init_completion(&master->xfer_completion); if (!master->max_dma_len) @@ -2237,6 +2243,35 @@ int spi_async_locked(struct spi_device *spi, struct spi_message *message) EXPORT_SYMBOL_GPL(spi_async_locked); +int spi_mtd_mmap_read(struct spi_device *spi, loff_t from, size_t len, + size_t *retlen, u_char *buf, u8 read_opcode, + u8 addr_width, u8 dummy_bytes) + +{ + struct spi_master *master = spi->master; + int ret; + + if (master->auto_runtime_pm) { + ret = pm_runtime_get_sync(master->dev.parent); + if (ret < 0) { + dev_err(&master->dev, "Failed to power device: %d\n", + ret); + goto err; + } + } + mutex_lock(&master->mmap_lock_mutex); + ret = master->spi_mtd_mmap_read(spi, from, len, retlen, buf, + read_opcode, addr_width, + dummy_bytes); + mutex_unlock(&master->mmap_lock_mutex); + if (master->auto_runtime_pm) + pm_runtime_put(master->dev.parent); + +err: + return ret; +} +EXPORT_SYMBOL_GPL(spi_mtd_mmap_read); + /*-------------------------------------------------------------------------*/ /* Utility methods for SPI master protocol drivers, layered on diff --git a/include/linux/spi/spi.h b/include/linux/spi/spi.h index 6b00f18f5e6b..0a6d8ad57357 100644 --- a/include/linux/spi/spi.h +++ b/include/linux/spi/spi.h @@ -297,6 +297,7 @@ static inline void spi_unregister_driver(struct spi_driver *sdrv) * @flags: other constraints relevant to this driver * @bus_lock_spinlock: spinlock for SPI bus locking * @bus_lock_mutex: mutex for SPI bus locking + * @mmap_lock_mutex: mutex for locking SPI bus when mmap transfer is on. * @bus_lock_flag: indicates that the SPI bus is locked for exclusive use * @setup: updates the device mode and clocking records used by a * device's SPI controller; protocol code may call this. This @@ -353,6 +354,11 @@ static inline void spi_unregister_driver(struct spi_driver *sdrv) * @handle_err: the subsystem calls the driver to handle an error that occurs * in the generic implementation of transfer_one_message(). * @unprepare_message: undo any work done by prepare_message(). + * @spi_mtd_mmap_read: some spi-controller hardwares provide memory. + * Flash drivers (like m25p80) can request memory + * mapped read via this method. This interface + * should only be used by mtd flashes and cannot be + * used by other spi devices. * @cs_gpios: Array of GPIOs to use as chip select lines; one per CS * number. Any individual value may be -ENOENT for CS lines that * are not GPIOs (driven by the SPI controller itself). @@ -420,6 +426,8 @@ struct spi_master { /* lock and mutex for SPI bus locking */ spinlock_t bus_lock_spinlock; struct mutex bus_lock_mutex; + /* mutex for SPI bus locking when mmap transfer is on */ + struct mutex mmap_lock_mutex; /* flag indicating that the SPI bus is locked for exclusive use */ bool bus_lock_flag; @@ -499,6 +507,11 @@ struct spi_master { struct spi_message *message); int (*unprepare_message)(struct spi_master *master, struct spi_message *message); + int (*spi_mtd_mmap_read)(struct spi_device *spi, + loff_t from, size_t len, + size_t *retlen, u_char *buf, + u8 read_opcode, u8 addr_width, + u8 dummy_bytes); /* * These hooks are for drivers that use a generic implementation @@ -985,6 +998,16 @@ static inline ssize_t spi_w8r16be(struct spi_device *spi, u8 cmd) return be16_to_cpu(result); } +/* SPI core interface for memory mapped read support */ +static inline bool spi_mmap_read_supported(struct spi_device *spi) +{ + return spi->master->spi_mtd_mmap_read ? true : false; +} + +int spi_mtd_mmap_read(struct spi_device *spi, loff_t from, size_t len, + size_t *retlen, u_char *buf, u8 read_opcode, + u8 addr_width, u8 dummy_bytes); + /*---------------------------------------------------------------------------*/ /*
In addition to providing direct access to SPI bus, some spi controller hardwares (like ti-qspi) provide special memory mapped port to accesses SPI flash devices in order to increase read performance. This means the controller can automatically send the SPI signals required to read data from the SPI flash device. For this, spi controller needs to know flash specific information like read command to use, dummy bytes and address width. Once these settings are populated in hardware registers, any read accesses to flash's memory map region(SoC specific) through memcpy (or mem-to mem DMA copy) will be handled by controller hardware. The hardware will automatically generate SPI signals required to read data from flash and present it to CPU/DMA. Introduce spi_mtd_mmap_read() interface to support memory mapped read over SPI flash devices. SPI master drivers can implement this callback to support memory mapped read interfaces. m25p80 flash driver and other flash drivers can call this to request memory mapped read. The interface should only be used MTD flashes and cannot be used with other SPI devices. Signed-off-by: Vignesh R <vigneshr@ti.com> --- drivers/spi/spi.c | 35 +++++++++++++++++++++++++++++++++++ include/linux/spi/spi.h | 23 +++++++++++++++++++++++ 2 files changed, 58 insertions(+)