Message ID | 1614599771-33629-1-git-send-email-f.fangjian@huawei.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | spi: Add HiSilicon SPI controller driver support | expand |
On Mon, Mar 01, 2021 at 07:56:11PM +0800, Jay Fang wrote: > This driver supports SPI Controller for HiSilicon Kunpeng SOCs. This > driver supports SPI operations using FIFO mode of transfer. > +HISILICON SPI Controller Driver > +M: Jay Fang <f.fangjian@huawei.com> > +L: linux-spi@vger.kernel.org > +S: Maintained > +W: http://www.hisilicon.com > +F: drivers/spi/spi-hisi.c Please give this a more specific name, the commit message already says this is for the Kunpeng SoCs but HiSilicon has other products, or may choose to use a different IP in some future Kunpeng part for that matter. > obj-$(CONFIG_SPI_GPIO) += spi-gpio.o > obj-$(CONFIG_SPI_HISI_SFC_V3XX) += spi-hisi-sfc-v3xx.o > +obj-$(CONFIG_SPI_HISI) += spi-hisi.o Please keep the Kconfig and Makefile sorted. > +++ b/drivers/spi/spi-hisi.c > @@ -0,0 +1,573 @@ > +// SPDX-License-Identifier: GPL-2.0-only > +/* Please make the entire comment a C++ one so things look more intentional. > + * HiSilicon SPI Controller Driver (refer spi-dw-core.c) This comment suggests that this is a variation of the DesignWare controller if that is the case please extend that driver rather than adding a totally new one. > +/* Disable IRQ bits */ > +static void hisi_spi_mask_intr(struct hisi_spi *hs, u32 mask) > +{ > + u32 new_mask; > + > + new_mask = readl(hs->regs + HISI_SPI_IMR) | mask; > + writel(new_mask, hs->regs + HISI_SPI_IMR); > +} This is a read/modify/write cycle and appears to be called from at least process and interrupt context but I'm not seeing anything that stops two different callers of it or the matching unmask function from running at the same time. > + while (hisi_spi_rx_not_empty(hs) && max--) { > + rxw = readl(hs->regs + HISI_SPI_DOUT); > + /* Check the transfer's original "rx" is not null */ > + if (hs->rx) { > + switch (hs->n_bytes) { > + case HISI_SPI_N_BYTES_U8: > + *(u8 *)(hs->rx) = rxw; > + break; > + case HISI_SPI_N_BYTES_U16: > + *(u16 *)(hs->rx) = rxw; > + break; > + case HISI_SPI_N_BYTES_U32: > + *(u32 *)(hs->rx) = rxw; > + break; > + } > + hs->rx += hs->n_bytes; > + } > + --hs->rx_len; You can probably get better performance by running some of the transfers in 32 bit or 16 bit mode, no need to do that to merge the driver though. Also are you sure that the length is tracked properly for things that aren't bytes? > +static irqreturn_t hisi_spi_handle_transfer(struct hisi_spi *hs, > + u32 irq_status) > +{ > + struct spi_controller *master = hs->master; > + > + /* Error handling */ > + if (irq_status & ISR_RXOF) { > + dev_err(&master->dev, "%s\n", > + "interrupt_transfer: fifo overflow"); There is no need to use the %s here, it just makes the display more confusing. > +static irqreturn_t hisi_spi_irq(int irq, void *dev_id) > +{ > + struct spi_controller *master = dev_id; > + struct hisi_spi *hs = spi_controller_get_devdata(master); > + u32 irq_status = readl(hs->regs + HISI_SPI_ISR) & ISR_MASK; > + > + if (!irq_status) > + return IRQ_NONE; > + > + if (!master->cur_msg) { > + hisi_spi_mask_intr(hs, IMR_MASK); > + return IRQ_HANDLED; > + } > + > + return hisi_spi_handle_transfer(hs, irq_status); It's probably clearer to just merge the two bits of this interrupt handler here, it's a bit confusing that they're separate functions and there's only the one caller of _handle_transfer(). > +static int hisi_spi_transfer_one(struct spi_controller *master, > + struct spi_device *spi, struct spi_transfer *transfer) > +{ > + struct hisi_spi *hs = spi_controller_get_devdata(master); > + > + hs->n_bytes = hisi_spi_n_bytes(transfer); > + hs->tx = (void *)transfer->tx_buf; If there's a need to cast to void * something is very wrong here. > + hs->tx_len = transfer->len / hs->n_bytes; > + hs->rx = transfer->rx_buf; > + hs->rx_len = hs->tx_len; > + > + /* Ensure the data above is visible for all CPUs */ > + smp_mb(); This memory barrier seems worrying... are you *sure* this is the best way to sync, and that the sync is best done here if it is needed rather than after everything else is set up? > + > + /* Disable is needed to deal with transfer timeout */ > + hisi_spi_disable(hs); > + > + hisi_spi_flush_fifo(hs); > + hisi_spi_update_cr(hs, spi, transfer); Especially given this, if there may be some left over operations going on elsewhere might there be races due to that? I'm also wondering if it's faster to just reset the controller as is done in some error handling paths. > + ret = devm_request_irq(dev, hs->irq, hisi_spi_irq, IRQF_SHARED, > + dev_name(dev), master); > + if (ret < 0) { > + dev_err(dev, "failed to get IRQ=%d, ret=%d\n", hs->irq, ret); > + return ret; > + } This will free the IRQ *after* the controller is unregistered, it's better to manually free the interrupt > + ret = hisi_spi_add_host(&pdev->dev, hs); > + if (ret) > + return ret; I don't see much value in splitting this out from the main probe function, again it's just making the code a bit more complex.
CC:linux-spi Dear Mark, Sorry, there are some format problems in the earlier reply email, please ignore it. Please reply based on this email. Thanks Jay On 2021/3/1 21:54, Mark Brown wrote: > On Mon, Mar 01, 2021 at 07:56:11PM +0800, Jay Fang wrote: > >> This driver supports SPI Controller for HiSilicon Kunpeng SOCs. This >> driver supports SPI operations using FIFO mode of transfer. > >> +HISILICON SPI Controller Driver >> +M: Jay Fang <f.fangjian@huawei.com> >> +L: linux-spi@vger.kernel.org >> +S: Maintained >> +W: http://www.hisilicon.com >> +F: drivers/spi/spi-hisi.c > > Please give this a more specific name, the commit message already says > this is for the Kunpeng SoCs but HiSilicon has other products, or may > choose to use a different IP in some future Kunpeng part for that > matter. Will fix. > >> obj-$(CONFIG_SPI_GPIO) += spi-gpio.o >> obj-$(CONFIG_SPI_HISI_SFC_V3XX) += spi-hisi-sfc-v3xx.o >> +obj-$(CONFIG_SPI_HISI) += spi-hisi.o > > Please keep the Kconfig and Makefile sorted. > Will fix. >> +++ b/drivers/spi/spi-hisi.c >> @@ -0,0 +1,573 @@ >> +// SPDX-License-Identifier: GPL-2.0-only >> +/* > > Please make the entire comment a C++ one so things look more > intentional. > Will fix. >> + * HiSilicon SPI Controller Driver (refer spi-dw-core.c) > > This comment suggests that this is a variation of the DesignWare > controller if that is the case please extend that driver rather than > adding a totally new one. > The hardware processing flow code is very different, so it would be unrealistic to extend one driver to achieve compatibility with two different IPs. >> +/* Disable IRQ bits */ >> +static void hisi_spi_mask_intr(struct hisi_spi *hs, u32 mask) >> +{ >> + u32 new_mask; >> + >> + new_mask = readl(hs->regs + HISI_SPI_IMR) | mask; >> + writel(new_mask, hs->regs + HISI_SPI_IMR); >> +} > > This is a read/modify/write cycle and appears to be called from at least > process and interrupt context but I'm not seeing anything that stops two > different callers of it or the matching unmask function from running at > the same time. > Those mask/unmask will not be called at the same time from process and interrupt context. In process context, unmask will be called after SPI controller be Disable and Flush (interrupt handing has ended). >> + while (hisi_spi_rx_not_empty(hs) && max--) { >> + rxw = readl(hs->regs + HISI_SPI_DOUT); >> + /* Check the transfer's original "rx" is not null */ >> + if (hs->rx) { >> + switch (hs->n_bytes) { >> + case HISI_SPI_N_BYTES_U8: >> + *(u8 *)(hs->rx) = rxw; >> + break; >> + case HISI_SPI_N_BYTES_U16: >> + *(u16 *)(hs->rx) = rxw; >> + break; >> + case HISI_SPI_N_BYTES_U32: >> + *(u32 *)(hs->rx) = rxw; >> + break; >> + } >> + hs->rx += hs->n_bytes; >> + } >> + --hs->rx_len; > > You can probably get better performance by running some of the transfers > in 32 bit or 16 bit mode, no need to do that to merge the driver though. > Also are you sure that the length is tracked properly for things that > aren't bytes? > hs->tx_len = transfer->len / hs->n_bytes; As the code above, the length is not in bytes. >> +static irqreturn_t hisi_spi_handle_transfer(struct hisi_spi *hs, >> + u32 irq_status) >> +{ >> + struct spi_controller *master = hs->master; >> + >> + /* Error handling */ >> + if (irq_status & ISR_RXOF) { >> + dev_err(&master->dev, "%s\n", >> + "interrupt_transfer: fifo overflow"); > > There is no need to use the %s here, it just makes the display more > confusing. > Will change. >> +static irqreturn_t hisi_spi_irq(int irq, void *dev_id) >> +{ >> + struct spi_controller *master = dev_id; >> + struct hisi_spi *hs = spi_controller_get_devdata(master); >> + u32 irq_status = readl(hs->regs + HISI_SPI_ISR) & ISR_MASK; >> + >> + if (!irq_status) >> + return IRQ_NONE; >> + >> + if (!master->cur_msg) { >> + hisi_spi_mask_intr(hs, IMR_MASK); >> + return IRQ_HANDLED; >> + } >> + >> + return hisi_spi_handle_transfer(hs, irq_status); > > It's probably clearer to just merge the two bits of this interrupt > handler here, it's a bit confusing that they're separate functions and > there's only the one caller of _handle_transfer(). > Will change. >> +static int hisi_spi_transfer_one(struct spi_controller *master, >> + struct spi_device *spi, struct spi_transfer *transfer) >> +{ >> + struct hisi_spi *hs = spi_controller_get_devdata(master); >> + >> + hs->n_bytes = hisi_spi_n_bytes(transfer); >> + hs->tx = (void *)transfer->tx_buf; > > If there's a need to cast to void * something is very wrong here. > Yes, fix compile warning. >> + hs->tx_len = transfer->len / hs->n_bytes; >> + hs->rx = transfer->rx_buf; >> + hs->rx_len = hs->tx_len; >> + >> + /* Ensure the data above is visible for all CPUs */ >> + smp_mb(); > > This memory barrier seems worrying... are you *sure* this is the best > way to sync, and that the sync is best done here if it is needed rather > than after everything else is set up? > The commit 0b6bfad ("spi: spi-dw: Remove extraneous locking") explains why memory barrier is needed here. And put it here to make it easier to understand. >> + >> + /* Disable is needed to deal with transfer timeout */ >> + hisi_spi_disable(hs); >> + >> + hisi_spi_flush_fifo(hs); >> + hisi_spi_update_cr(hs, spi, transfer); > > Especially given this, if there may be some left over operations going > on elsewhere might there be races due to that? I'm also wondering if > it's faster to just reset the controller as is done in some error > handling paths. Will fix this. Disable and Flush will be done before fill hisi_spi's transfer info. > >> + ret = devm_request_irq(dev, hs->irq, hisi_spi_irq, IRQF_SHARED, >> + dev_name(dev), master); >> + if (ret < 0) { >> + dev_err(dev, "failed to get IRQ=%d, ret=%d\n", hs->irq, ret); >> + return ret; >> + } > > This will free the IRQ *after* the controller is unregistered, it's > better to manually free the interrupt > Ok >> + ret = hisi_spi_add_host(&pdev->dev, hs); >> + if (ret) >> + return ret; > > I don't see much value in splitting this out from the main probe > function, again it's just making the code a bit more complex. Ok >
On Thu, Mar 04, 2021 at 10:54:40AM +0800, Fangjian (Jay) wrote: > On 2021/3/1 21:54, Mark Brown wrote: > > On Mon, Mar 01, 2021 at 07:56:11PM +0800, Jay Fang wrote: > > > +/* Disable IRQ bits */ > > > +static void hisi_spi_mask_intr(struct hisi_spi *hs, u32 mask) > > > +{ > > > + u32 new_mask; > > > + > > > + new_mask = readl(hs->regs + HISI_SPI_IMR) | mask; > > > + writel(new_mask, hs->regs + HISI_SPI_IMR); > > > +} > > This is a read/modify/write cycle and appears to be called from at least > > process and interrupt context but I'm not seeing anything that stops two > > different callers of it or the matching unmask function from running at > > the same time. > Those mask/unmask will not be called at the same time from process and > interrupt context. In process context, unmask will be called after SPI > controller be Disable and Flush (interrupt handing has ended). Given that this is disabling the interrupt that doesn't sound like it's going to be entirely robust - if we need to disable the interrupt presumably there's some chance it might fire. > > > + struct hisi_spi *hs = spi_controller_get_devdata(master); > > > + > > > + hs->n_bytes = hisi_spi_n_bytes(transfer); > > > + hs->tx = (void *)transfer->tx_buf; > > If there's a need to cast to void * something is very wrong here. > Yes, fix compile warning. This cast just masks whatever the problem is, if the compiler is complaining about using a void pointer it's spotted an issue. > > > + /* Ensure the data above is visible for all CPUs */ > > > + smp_mb(); > > This memory barrier seems worrying... are you *sure* this is the best > > way to sync, and that the sync is best done here if it is needed rather > > than after everything else is set up? > The commit 0b6bfad ("spi: spi-dw: Remove extraneous locking") explains > why memory barrier is needed here. And put it here to make it easier to > understand. The reader of this code won't have any kind of pointer to that commit, this needs to be clearer.
On Mon, Mar 01, 2021 at 07:56:11PM +0800, Jay Fang wrote: > +static int hisi_spi_setup(struct spi_device *spi) > +{ > + struct hisi_spi *hs = spi_controller_get_devdata(spi->controller); > + struct hisi_chip_data *chip; > + > + /* Only alloc on first setup */ > + chip = spi_get_ctldata(spi); > + if (!chip) { > + chip = kzalloc(sizeof(*chip), GFP_KERNEL); > + if (!chip) > + return -ENOMEM; > + spi_set_ctldata(spi, chip); > + } [...] > + master = devm_spi_alloc_master(dev, 0); It would seem simpler to just pass "sizeof struct hisi_chip_data" to devm_spi_alloc_master(), thus obviating the need to allocate it on first incocation of the ->setup hook. Thanks, Lukas
On Mon, Mar 01, 2021 at 01:54:05PM +0000, Mark Brown wrote: > On Mon, Mar 01, 2021 at 07:56:11PM +0800, Jay Fang wrote: > > + ret = devm_request_irq(dev, hs->irq, hisi_spi_irq, IRQF_SHARED, > > + dev_name(dev), master); > > + if (ret < 0) { > > + dev_err(dev, "failed to get IRQ=%d, ret=%d\n", hs->irq, ret); > > + return ret; > > + } > > This will free the IRQ *after* the controller is unregistered, it's > better to manually free the interrupt Transfers may still be ongoing until spi_unregister_controller() returns. (It's called from devres_release_all() in this case.) Since the IRQ is presumably necessary to handle those transfers, freeing the IRQ after unregistering is actually correct. So the code looks fine in principle. However, because the IRQ is requested with IRQF_SHARED, the handler may be invoked at any time, even after the controller has been unregistered. It is therefore necessary to quiesce the SPI controller's interrupt on unregistering and it is also necessary to check in the IRQ handler whether an interrupt is actually pending (and bail out if not). Thanks, Lukas
On 2021/3/7 22:36, Lukas Wunner wrote: > On Mon, Mar 01, 2021 at 07:56:11PM +0800, Jay Fang wrote: >> +static int hisi_spi_setup(struct spi_device *spi) >> +{ >> + struct hisi_spi *hs = spi_controller_get_devdata(spi->controller); >> + struct hisi_chip_data *chip; >> + >> + /* Only alloc on first setup */ >> + chip = spi_get_ctldata(spi); >> + if (!chip) { >> + chip = kzalloc(sizeof(*chip), GFP_KERNEL); >> + if (!chip) >> + return -ENOMEM; >> + spi_set_ctldata(spi, chip); >> + } > [...] >> + master = devm_spi_alloc_master(dev, 0); > > It would seem simpler to just pass "sizeof struct hisi_chip_data" > to devm_spi_alloc_master(), thus obviating the need to allocate > it on first incocation of the ->setup hook. > Hi Lukas, Thanks for your review. "struct hisi_chip_data" is the private data of per-spi device, not the spi controller. So I think this method is not suitable here. Maybe we can allocate "struct hisi_spi" in this way. Thanks Jay
On Mon, Mar 08, 2021 at 11:57:17AM +0800, Jay Fang wrote: > On 2021/3/7 22:36, Lukas Wunner wrote: > > On Mon, Mar 01, 2021 at 07:56:11PM +0800, Jay Fang wrote: > > > +static int hisi_spi_setup(struct spi_device *spi) > > > +{ > > > + struct hisi_spi *hs = spi_controller_get_devdata(spi->controller); > > > + struct hisi_chip_data *chip; > > > + > > > + /* Only alloc on first setup */ > > > + chip = spi_get_ctldata(spi); > > > + if (!chip) { > > > + chip = kzalloc(sizeof(*chip), GFP_KERNEL); > > > + if (!chip) > > > + return -ENOMEM; > > > + spi_set_ctldata(spi, chip); > > > + } > > [...] > > > + master = devm_spi_alloc_master(dev, 0); > > > > It would seem simpler to just pass "sizeof struct hisi_chip_data" > > to devm_spi_alloc_master(), thus obviating the need to allocate > > it on first invocation of the ->setup hook. > > Thanks for your review. "struct hisi_chip_data" is the private data > of per-spi device, not the spi controller. So I think this method > is not suitable here. Maybe we can allocate "struct hisi_spi" in > this way. Ah, I was mistaken then. Yes, allocating struct hisi_spi with devm_spi_alloc_master() would probably make sense. Thanks, Lukas
Dear Mark, On 2021/3/4 20:34, Mark Brown wrote: > On Thu, Mar 04, 2021 at 10:54:40AM +0800, Fangjian (Jay) wrote: >> On 2021/3/1 21:54, Mark Brown wrote: >>> On Mon, Mar 01, 2021 at 07:56:11PM +0800, Jay Fang wrote: > >>>> +/* Disable IRQ bits */ >>>> +static void hisi_spi_mask_intr(struct hisi_spi *hs, u32 mask) >>>> +{ >>>> + u32 new_mask; >>>> + >>>> + new_mask = readl(hs->regs + HISI_SPI_IMR) | mask; >>>> + writel(new_mask, hs->regs + HISI_SPI_IMR); >>>> +} > >>> This is a read/modify/write cycle and appears to be called from at least >>> process and interrupt context but I'm not seeing anything that stops two >>> different callers of it or the matching unmask function from running at >>> the same time. > >> Those mask/unmask will not be called at the same time from process and >> interrupt context. In process context, unmask will be called after SPI >> controller be Disable and Flush (interrupt handing has ended). > > Given that this is disabling the interrupt that doesn't sound like it's > going to be entirely robust - if we need to disable the interrupt > presumably there's some chance it might fire. > + > + /* Disable is needed to deal with transfer timeout */ > + hisi_spi_disable(hs); > + > + hisi_spi_flush_fifo(hs); > + hisi_spi_update_cr(hs, spi, transfer); Especially given this, if there may be some left over operations going on elsewhere might there be races due to that? I'm also wondering if it's faster to just reset the controller as is done in some error handling paths. (Mark) As the previous review said, would it be better to reset controller and waiting for interrupt to complete in error handling paths? As shown in the code below. static void hisi_spi_handle_err(struct spi_controller *master, struct spi_message *msg) { struct hisi_spi *hs = spi_controller_get_devdata(master); hisi_spi_reset(hs); hisi_spi_flush_fifo(hs); /* * Wait for interrupt handler that is * already in timeout to complete. */ synchronize_irq(hs->irq); } This ensures that the new transfer will not start until the previous transfer is completed. And we don't need to disable the controller every time before data transfer. Thanks Jay > >>>> + struct hisi_spi *hs = spi_controller_get_devdata(master); >>>> + >>>> + hs->n_bytes = hisi_spi_n_bytes(transfer); >>>> + hs->tx = (void *)transfer->tx_buf; >>> If there's a need to cast to void * something is very wrong here. > >> Yes, fix compile warning. > > This cast just masks whatever the problem is, if the compiler is > complaining about using a void pointer it's spotted an issue. Will fix. > >>>> + /* Ensure the data above is visible for all CPUs */ >>>> + smp_mb(); > >>> This memory barrier seems worrying... are you *sure* this is the best >>> way to sync, and that the sync is best done here if it is needed rather >>> than after everything else is set up? > >> The commit 0b6bfad ("spi: spi-dw: Remove extraneous locking") explains >> why memory barrier is needed here. And put it here to make it easier to >> understand. > > The reader of this code won't have any kind of pointer to that commit, > this needs to be clearer. Will fix. >
On Sun, Mar 07, 2021 at 03:43:13PM +0100, Lukas Wunner wrote: > On Mon, Mar 01, 2021 at 01:54:05PM +0000, Mark Brown wrote: > > On Mon, Mar 01, 2021 at 07:56:11PM +0800, Jay Fang wrote: > > > + ret = devm_request_irq(dev, hs->irq, hisi_spi_irq, IRQF_SHARED, > > > + dev_name(dev), master); > > > + if (ret < 0) { > > > + dev_err(dev, "failed to get IRQ=%d, ret=%d\n", hs->irq, ret); > > > + return ret; > > > + } > > This will free the IRQ *after* the controller is unregistered, it's > > better to manually free the interrupt > Transfers may still be ongoing until spi_unregister_controller() returns. > (It's called from devres_release_all() in this case.) Since the IRQ is > presumably necessary to handle those transfers, freeing the IRQ after > unregistering is actually correct. So the code looks fine in principle. Well, it should just time out which is annoying. > However, because the IRQ is requested with IRQF_SHARED, the handler may > be invoked at any time, even after the controller has been unregistered. > It is therefore necessary to quiesce the SPI controller's interrupt on > unregistering and it is also necessary to check in the IRQ handler whether > an interrupt is actually pending (and bail out if not). It's also important and even more of a concern that even if there is a valid interrupt the handler doesn't try to use structures that might have been deallocated before the handler uses it as this controller does, that will segfault which is more serious.
On Mon, Mar 08, 2021 at 02:11:01PM +0000, Mark Brown wrote: > On Sun, Mar 07, 2021 at 03:43:13PM +0100, Lukas Wunner wrote: > > Transfers may still be ongoing until spi_unregister_controller() returns. > > (It's called from devres_release_all() in this case.) Since the IRQ is > > presumably necessary to handle those transfers, freeing the IRQ after > > unregistering is actually correct. So the code looks fine in principle. > > However, because the IRQ is requested with IRQF_SHARED, the handler may > > be invoked at any time, even after the controller has been unregistered. > > It is therefore necessary to quiesce the SPI controller's interrupt on > > unregistering and it is also necessary to check in the IRQ handler whether > > an interrupt is actually pending (and bail out if not). > > It's also important and even more of a concern that even if there is a > valid interrupt the handler doesn't try to use structures that might > have been deallocated before the handler uses it as this controller > does, that will segfault which is more serious. At least struct spi_controller and struct hisi_spi are allocated with devm_*() before the call to devm_request_irq(), hence those two are always accessible from the IRQ handler AFAICS. Thanks, Lukas
On Mon, Mar 08, 2021 at 07:18:39PM +0100, Lukas Wunner wrote: > On Mon, Mar 08, 2021 at 02:11:01PM +0000, Mark Brown wrote: > > It's also important and even more of a concern that even if there is a > > valid interrupt the handler doesn't try to use structures that might > > have been deallocated before the handler uses it as this controller > > does, that will segfault which is more serious. > At least struct spi_controller and struct hisi_spi are allocated with > devm_*() before the call to devm_request_irq(), hence those two > are always accessible from the IRQ handler AFAICS. For stuff like this I'd rather the code were obviously correct rather than having to think about it, dereferencing the struct is the most obvious thing but there's also the possibility that we cause something else to happen which explodes on us, or that someone later modifies the code in a way that introduces an issue. It does look safe right now but the whole implicitly managed asynchronous code thing always makes me nervous.
On 2021/3/7 22:43, Lukas Wunner wrote: > On Mon, Mar 01, 2021 at 01:54:05PM +0000, Mark Brown wrote: >> On Mon, Mar 01, 2021 at 07:56:11PM +0800, Jay Fang wrote: >>> + ret = devm_request_irq(dev, hs->irq, hisi_spi_irq, IRQF_SHARED, >>> + dev_name(dev), master); >>> + if (ret < 0) { >>> + dev_err(dev, "failed to get IRQ=%d, ret=%d\n", hs->irq, ret); >>> + return ret; >>> + } >> >> This will free the IRQ *after* the controller is unregistered, it's >> better to manually free the interrupt > > Transfers may still be ongoing until spi_unregister_controller() returns. > (It's called from devres_release_all() in this case.) Since the IRQ is > presumably necessary to handle those transfers, freeing the IRQ after > unregistering is actually correct. So the code looks fine in principle. > > However, because the IRQ is requested with IRQF_SHARED, the handler may > be invoked at any time, even after the controller has been unregistered. > It is therefore necessary to quiesce the SPI controller's interrupt on > unregistering and it is also necessary to check in the IRQ handler whether > an interrupt is actually pending (and bail out if not). > Thanks for your review. I will consider not using IRQF_SHARED IRQ here. Thanks Jay
diff --git a/MAINTAINERS b/MAINTAINERS index 546aa66..5769049 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -8080,6 +8080,13 @@ F: drivers/crypto/hisilicon/sec2/sec_crypto.c F: drivers/crypto/hisilicon/sec2/sec_crypto.h F: drivers/crypto/hisilicon/sec2/sec_main.c +HISILICON SPI Controller Driver +M: Jay Fang <f.fangjian@huawei.com> +L: linux-spi@vger.kernel.org +S: Maintained +W: http://www.hisilicon.com +F: drivers/spi/spi-hisi.c + HISILICON STAGING DRIVERS FOR HIKEY 960/970 M: Mauro Carvalho Chehab <mchehab+huawei@kernel.org> L: devel@driverdev.osuosl.org diff --git a/drivers/spi/Kconfig b/drivers/spi/Kconfig index aadaea0..94c4b57 100644 --- a/drivers/spi/Kconfig +++ b/drivers/spi/Kconfig @@ -339,6 +339,16 @@ config SPI_FSL_QUADSPI This controller does not support generic SPI messages. It only supports the high-level SPI memory interface. +config SPI_HISI + tristate "HiSilicon SPI controller" + depends on (ARM64 && ACPI) || COMPILE_TEST + help + Say Y here if you want have SPI controller support available + on the HiSilicon Kunpeng SOCs. + + This driver can also be built as a module. If so, the module + will be called spi-hisi. + config SPI_HISI_SFC_V3XX tristate "HiSilicon SPI NOR Flash Controller for Hi16XX chipsets" depends on (ARM64 && ACPI) || COMPILE_TEST diff --git a/drivers/spi/Makefile b/drivers/spi/Makefile index 6fea582..af3029d 100644 --- a/drivers/spi/Makefile +++ b/drivers/spi/Makefile @@ -55,6 +55,7 @@ obj-$(CONFIG_SPI_FSL_QUADSPI) += spi-fsl-qspi.o obj-$(CONFIG_SPI_FSL_SPI) += spi-fsl-spi.o obj-$(CONFIG_SPI_GPIO) += spi-gpio.o obj-$(CONFIG_SPI_HISI_SFC_V3XX) += spi-hisi-sfc-v3xx.o +obj-$(CONFIG_SPI_HISI) += spi-hisi.o obj-$(CONFIG_SPI_IMG_SPFI) += spi-img-spfi.o obj-$(CONFIG_SPI_IMX) += spi-imx.o obj-$(CONFIG_SPI_LANTIQ_SSC) += spi-lantiq-ssc.o diff --git a/drivers/spi/spi-hisi.c b/drivers/spi/spi-hisi.c new file mode 100644 index 00000000..5d48bbf --- /dev/null +++ b/drivers/spi/spi-hisi.c @@ -0,0 +1,573 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * HiSilicon SPI Controller Driver (refer spi-dw-core.c) + * + * Copyright (c) 2021, HiSilicon Corporation. + */ + +#include <linux/acpi.h> +#include <linux/bitfield.h> +#include <linux/delay.h> +#include <linux/err.h> +#include <linux/interrupt.h> +#include <linux/module.h> +#include <linux/property.h> +#include <linux/platform_device.h> +#include <linux/slab.h> +#include <linux/spi/spi.h> + +/* Register offsets */ +#define HISI_SPI_CSCR 0x00 /* cs control register */ +#define HISI_SPI_CR 0x04 /* spi common control register */ +#define HISI_SPI_ENR 0x08 /* spi enable register */ +#define HISI_SPI_FIFOC 0x0c /* fifo level control register */ +#define HISI_SPI_IMR 0x10 /* interrupt mask register */ +#define HISI_SPI_DIN 0x14 /* data in register */ +#define HISI_SPI_DOUT 0x18 /* data out register */ +#define HISI_SPI_SR 0x1c /* status register */ +#define HISI_SPI_RISR 0x20 /* raw interrupt status register */ +#define HISI_SPI_ISR 0x24 /* interrupt status register */ +#define HISI_SPI_ICR 0x28 /* interrupt clear register */ +#define HISI_SPI_VERSION 0xfe0 /* version register */ + +/* Bit fields in HISI_SPI_CR */ +#define CR_LOOP_MASK GENMASK(1, 1) +#define CR_CPOL_MASK GENMASK(2, 2) +#define CR_CPHA_MASK GENMASK(3, 3) +#define CR_DIV_PRE_MASK GENMASK(11, 4) +#define CR_DIV_POST_MASK GENMASK(19, 12) +#define CR_BPW_MASK GENMASK(24, 20) +#define CR_SPD_MODE_MASK GENMASK(25, 25) + +/* Bit fields in HISI_SPI_FIFOC */ +#define FIFOC_TX_MASK GENMASK(5, 3) +#define FIFOC_RX_MASK GENMASK(11, 9) + +/* Bit fields in HISI_SPI_IMR, 4 bits */ +#define IMR_RXOF BIT(0) /* Receive Overflow */ +#define IMR_RXTO BIT(1) /* Receive Timeout */ +#define IMR_RX BIT(2) /* Receive */ +#define IMR_TX BIT(3) /* Transmit */ +#define IMR_MASK (IMR_RXOF | IMR_RXTO | IMR_RX | IMR_TX) + +/* Bit fields in HISI_SPI_SR, 5 bits */ +#define SR_TXE BIT(0) /* Transmit FIFO empty */ +#define SR_TXNF BIT(1) /* Transmit FIFO not full */ +#define SR_RXNE BIT(2) /* Receive FIFO not empty */ +#define SR_RXF BIT(3) /* Receive FIFO full */ +#define SR_BUSY BIT(4) /* Busy Flag */ + +/* Bit fields in HISI_SPI_ISR, 4 bits */ +#define ISR_RXOF BIT(0) /* Receive Overflow */ +#define ISR_RXTO BIT(1) /* Receive Timeout */ +#define ISR_RX BIT(2) /* Receive */ +#define ISR_TX BIT(3) /* Transmit */ +#define ISR_MASK (ISR_RXOF | ISR_RXTO | ISR_RX | ISR_TX) + +/* Bit fields in HISI_SPI_ICR, 2 bits */ +#define ICR_RXOF BIT(0) /* Receive Overflow */ +#define ICR_RXTO BIT(1) /* Receive Timeout */ +#define ICR_MASK (ICR_RXOF | ICR_RXTO) + +#define DIV_POST_MAX 0xFF +#define DIV_POST_MIN 0x00 +#define DIV_PRE_MAX 0xFE +#define DIV_PRE_MIN 0x02 +#define CLK_DIV_MAX ((1 + DIV_POST_MAX) * DIV_PRE_MAX) +#define CLK_DIV_MIN ((1 + DIV_POST_MIN) * DIV_PRE_MIN) + +#define DEFAULT_NUM_CS 1 + +enum hisi_spi_rx_level_trig { + HISI_SPI_RX_1, + HISI_SPI_RX_4, + HISI_SPI_RX_8, + HISI_SPI_RX_16, + HISI_SPI_RX_32, + HISI_SPI_RX_64, + HISI_SPI_RX_128 +}; + +enum hisi_spi_tx_level_trig { + HISI_SPI_TX_1_OR_LESS, + HISI_SPI_TX_4_OR_LESS, + HISI_SPI_TX_8_OR_LESS, + HISI_SPI_TX_16_OR_LESS, + HISI_SPI_TX_32_OR_LESS, + HISI_SPI_TX_64_OR_LESS, + HISI_SPI_TX_128_OR_LESS +}; + +enum hisi_spi_frame_n_bytes { + HISI_SPI_N_BYTES_NULL, + HISI_SPI_N_BYTES_U8, + HISI_SPI_N_BYTES_U16, + HISI_SPI_N_BYTES_U32 = 4 +}; + +/* Slave spi_dev related */ +struct hisi_chip_data { + u32 cr; + u32 speed_hz; /* baud rate */ + u16 clk_div; /* baud rate divider */ + + /* clk_div = (1 + div_post) * div_pre */ + u8 div_post; /* value from 0 to 255 */ + u8 div_pre; /* value from 2 to 254 (even only!) */ +}; + +struct hisi_spi { + struct spi_controller *master; + + void __iomem *regs; + int irq; + u32 fifo_len; /* depth of the FIFO buffer */ + u32 max_freq; /* max bus freq supported */ + u16 bus_num; + u16 num_cs; /* supported slave numbers */ + + /* Current message transfer state info */ + void *tx; + unsigned int tx_len; + void *rx; + unsigned int rx_len; + u8 n_bytes; /* current is a 1/2/4 bytes op */ +}; + +static void hisi_spi_enable(struct hisi_spi *hs) +{ + writel(1, hs->regs + HISI_SPI_ENR); +} + +static void hisi_spi_disable(struct hisi_spi *hs) +{ + writel(0, hs->regs + HISI_SPI_ENR); +} + +/* Disable IRQ bits */ +static void hisi_spi_mask_intr(struct hisi_spi *hs, u32 mask) +{ + u32 new_mask; + + new_mask = readl(hs->regs + HISI_SPI_IMR) | mask; + writel(new_mask, hs->regs + HISI_SPI_IMR); +} + +/* Enable IRQ bits */ +static void hisi_spi_umask_intr(struct hisi_spi *hs, u32 mask) +{ + u32 new_mask; + + new_mask = readl(hs->regs + HISI_SPI_IMR) & ~mask; + writel(new_mask, hs->regs + HISI_SPI_IMR); +} + +/* Clear IRQ bits */ +static void hisi_spi_clear_intr(struct hisi_spi *hs, u32 clear) +{ + writel(clear, hs->regs + HISI_SPI_ICR); +} + +static u32 hisi_spi_busy(struct hisi_spi *hs) +{ + return readl(hs->regs + HISI_SPI_SR) & SR_BUSY; +} + +static u32 hisi_spi_rx_not_empty(struct hisi_spi *hs) +{ + return readl(hs->regs + HISI_SPI_SR) & SR_RXNE; +} + +static u32 hisi_spi_tx_not_full(struct hisi_spi *hs) +{ + return readl(hs->regs + HISI_SPI_SR) & SR_TXNF; +} + +/* Disable the controller and all interrupts */ +static void hisi_spi_reset(struct hisi_spi *hs) +{ + hisi_spi_disable(hs); + hisi_spi_mask_intr(hs, IMR_MASK); + hisi_spi_clear_intr(hs, ICR_MASK); +} + +static u8 hisi_spi_n_bytes(struct spi_transfer *transfer) +{ + if (transfer->bits_per_word <= 8) + return HISI_SPI_N_BYTES_U8; + else if (transfer->bits_per_word <= 16) + return HISI_SPI_N_BYTES_U16; + else + return HISI_SPI_N_BYTES_U32; +} + +static void hisi_spi_reader(struct hisi_spi *hs) +{ + u32 max = min_t(u32, hs->rx_len, hs->fifo_len); + u32 rxw; + + while (hisi_spi_rx_not_empty(hs) && max--) { + rxw = readl(hs->regs + HISI_SPI_DOUT); + /* Check the transfer's original "rx" is not null */ + if (hs->rx) { + switch (hs->n_bytes) { + case HISI_SPI_N_BYTES_U8: + *(u8 *)(hs->rx) = rxw; + break; + case HISI_SPI_N_BYTES_U16: + *(u16 *)(hs->rx) = rxw; + break; + case HISI_SPI_N_BYTES_U32: + *(u32 *)(hs->rx) = rxw; + break; + } + hs->rx += hs->n_bytes; + } + --hs->rx_len; + } +} + +static void hisi_spi_writer(struct hisi_spi *hs) +{ + u32 max = min_t(u32, hs->tx_len, hs->fifo_len); + u32 txw = 0; + + while (hisi_spi_tx_not_full(hs) && max--) { + /* Check the transfer's original "tx" is not null */ + if (hs->tx) { + switch (hs->n_bytes) { + case HISI_SPI_N_BYTES_U8: + txw = *(u8 *)(hs->tx); + break; + case HISI_SPI_N_BYTES_U16: + txw = *(u16 *)(hs->tx); + break; + case HISI_SPI_N_BYTES_U32: + txw = *(u32 *)(hs->tx); + break; + } + hs->tx += hs->n_bytes; + } + writel(txw, hs->regs + HISI_SPI_DIN); + --hs->tx_len; + } +} + +static void hisi_spi_flush_fifo(struct hisi_spi *hs) +{ + unsigned long limit = loops_per_jiffy << 1; + + do { + while (hisi_spi_rx_not_empty(hs)) + readl(hs->regs + HISI_SPI_DOUT); + } while (hisi_spi_busy(hs) && limit--); +} + +static void __hisi_calc_div_reg(struct hisi_chip_data *chip) +{ + chip->div_pre = DIV_PRE_MAX; + while (chip->div_pre >= DIV_PRE_MIN) { + if (chip->clk_div % chip->div_pre == 0) + break; + + chip->div_pre -= 2; + } + + if (chip->div_pre > chip->clk_div) + chip->div_pre = chip->clk_div; + + chip->div_post = (chip->clk_div / chip->div_pre) - 1; +} + +static u32 hisi_calc_effective_speed(struct hisi_spi *hs, + struct hisi_chip_data *chip, u32 speed_hz) +{ + u32 effective_speed; + + /* Note clock divider doesn't support odd numbers */ + chip->clk_div = DIV_ROUND_UP(hs->max_freq, speed_hz) + 1; + chip->clk_div &= 0xfffe; + if (chip->clk_div > CLK_DIV_MAX) + chip->clk_div = CLK_DIV_MAX; + + effective_speed = hs->max_freq / chip->clk_div; + if (chip->speed_hz != effective_speed) { + __hisi_calc_div_reg(chip); + chip->speed_hz = effective_speed; + } + + return effective_speed; +} + +static u32 hisi_spi_prepare_cr(struct hisi_spi *hs, + struct spi_device *spi) +{ + u32 cr; + + cr = FIELD_PREP(CR_SPD_MODE_MASK, 1); + + cr |= FIELD_PREP(CR_CPHA_MASK, (spi->mode & SPI_CPHA) ? 1 : 0); + cr |= FIELD_PREP(CR_CPOL_MASK, (spi->mode & SPI_CPOL) ? 1 : 0); + cr |= FIELD_PREP(CR_LOOP_MASK, (spi->mode & SPI_LOOP) ? 1 : 0); + + return cr; +} + +/* Handle per transfer options for speed and bpw */ +static void hisi_spi_update_cr(struct hisi_spi *hs, + struct spi_device *spi, struct spi_transfer *transfer) +{ + struct hisi_chip_data *chip = spi_get_ctldata(spi); + u32 cr = chip->cr; + + transfer->effective_speed_hz = + hisi_calc_effective_speed(hs, chip, transfer->speed_hz); + + cr |= FIELD_PREP(CR_DIV_PRE_MASK, chip->div_pre); + cr |= FIELD_PREP(CR_DIV_POST_MASK, chip->div_post); + cr |= FIELD_PREP(CR_BPW_MASK, transfer->bits_per_word - 1); + + writel(cr, hs->regs + HISI_SPI_CR); +} + +static void hisi_spi_hw_init(struct hisi_spi *hs) +{ + hisi_spi_reset(hs); + + /* FIFO default config */ + writel(FIELD_PREP(FIFOC_TX_MASK, HISI_SPI_TX_64_OR_LESS) | + FIELD_PREP(FIFOC_RX_MASK, HISI_SPI_RX_16), + hs->regs + HISI_SPI_FIFOC); + + hs->fifo_len = 256; +} + +static irqreturn_t hisi_spi_handle_transfer(struct hisi_spi *hs, + u32 irq_status) +{ + struct spi_controller *master = hs->master; + + /* Error handling */ + if (irq_status & ISR_RXOF) { + dev_err(&master->dev, "%s\n", + "interrupt_transfer: fifo overflow"); + master->cur_msg->status = -EIO; + goto finalize_transfer; + } + + /* + * Read data from the Rx FIFO every time we've got a chance + * executing this method. If there is nothing left to receive, + * clear irq status and terminate the procedure. + */ + hisi_spi_reader(hs); + if (!hs->rx_len) + goto finalize_transfer; + + /* + * Send data out if Tx FIFO IRQ is received. The Tx IRQ will be + * disabled after the data transmission is finished. + */ + if (irq_status & ISR_TX) { + hisi_spi_writer(hs); + + if (!hs->tx_len) + hisi_spi_mask_intr(hs, ISR_TX); + } + + return IRQ_HANDLED; + +finalize_transfer: + hisi_spi_reset(hs); + spi_finalize_current_transfer(master); + return IRQ_HANDLED; + +} + +static irqreturn_t hisi_spi_irq(int irq, void *dev_id) +{ + struct spi_controller *master = dev_id; + struct hisi_spi *hs = spi_controller_get_devdata(master); + u32 irq_status = readl(hs->regs + HISI_SPI_ISR) & ISR_MASK; + + if (!irq_status) + return IRQ_NONE; + + if (!master->cur_msg) { + hisi_spi_mask_intr(hs, IMR_MASK); + return IRQ_HANDLED; + } + + return hisi_spi_handle_transfer(hs, irq_status); +} + +static int hisi_spi_transfer_one(struct spi_controller *master, + struct spi_device *spi, struct spi_transfer *transfer) +{ + struct hisi_spi *hs = spi_controller_get_devdata(master); + + hs->n_bytes = hisi_spi_n_bytes(transfer); + hs->tx = (void *)transfer->tx_buf; + hs->tx_len = transfer->len / hs->n_bytes; + hs->rx = transfer->rx_buf; + hs->rx_len = hs->tx_len; + + /* Ensure the data above is visible for all CPUs */ + smp_mb(); + + /* Disable is needed to deal with transfer timeout */ + hisi_spi_disable(hs); + + hisi_spi_flush_fifo(hs); + hisi_spi_update_cr(hs, spi, transfer); + + /* Kickoff */ + hisi_spi_umask_intr(hs, IMR_MASK); + hisi_spi_enable(hs); + + return 1; +} + +static void hisi_spi_handle_err(struct spi_controller *master, + struct spi_message *msg) +{ + struct hisi_spi *hs = spi_controller_get_devdata(master); + + hisi_spi_reset(hs); +} + +static int hisi_spi_setup(struct spi_device *spi) +{ + struct hisi_spi *hs = spi_controller_get_devdata(spi->controller); + struct hisi_chip_data *chip; + + /* Only alloc on first setup */ + chip = spi_get_ctldata(spi); + if (!chip) { + chip = kzalloc(sizeof(*chip), GFP_KERNEL); + if (!chip) + return -ENOMEM; + spi_set_ctldata(spi, chip); + } + + chip->cr = hisi_spi_prepare_cr(hs, spi); + + return 0; +} + +static void hisi_spi_cleanup(struct spi_device *spi) +{ + struct hisi_chip_data *chip = spi_get_ctldata(spi); + + kfree(chip); + spi_set_ctldata(spi, NULL); +} + +static int hisi_spi_add_host(struct device *dev, struct hisi_spi *hs) +{ + struct spi_controller *master; + int ret; + + master = devm_spi_alloc_master(dev, 0); + if (!master) + return -ENOMEM; + + hs->master = master; + spi_controller_set_devdata(master, hs); + + hisi_spi_hw_init(hs); + + ret = devm_request_irq(dev, hs->irq, hisi_spi_irq, IRQF_SHARED, + dev_name(dev), master); + if (ret < 0) { + dev_err(dev, "failed to get IRQ=%d, ret=%d\n", hs->irq, ret); + return ret; + } + + master->use_gpio_descriptors = true; + master->mode_bits = SPI_CPOL | SPI_CPHA | SPI_CS_HIGH | SPI_LOOP; + master->bits_per_word_mask = SPI_BPW_RANGE_MASK(4, 32); + master->bus_num = hs->bus_num; + master->num_chipselect = hs->num_cs; + master->setup = hisi_spi_setup; + master->cleanup = hisi_spi_cleanup; + master->transfer_one = hisi_spi_transfer_one; + master->handle_err = hisi_spi_handle_err; + master->max_speed_hz = hs->max_freq; + master->dev.fwnode = dev->fwnode; + + ret = devm_spi_register_controller(dev, master); + if (ret) { + dev_err(&master->dev, + "failed to register spi master, ret=%d\n", ret); + return ret; + } + + return 0; +} + +static int hisi_spi_probe(struct platform_device *pdev) +{ + struct hisi_spi *hs; + u32 version; + int ret; + + hs = devm_kzalloc(&pdev->dev, sizeof(*hs), GFP_KERNEL); + if (!hs) + return -ENOMEM; + + hs->regs = devm_platform_ioremap_resource(pdev, 0); + if (IS_ERR(hs->regs)) + return PTR_ERR(hs->regs); + + hs->irq = platform_get_irq(pdev, 0); + if (hs->irq < 0) + return hs->irq; + + hs->bus_num = pdev->id; + + ret = device_property_read_u16(&pdev->dev, "num-cs", &hs->num_cs); + if (ret) + hs->num_cs = DEFAULT_NUM_CS; + + /* Specify maximum SPI clocking speed (master only) by firmware */ + ret = device_property_read_u32(&pdev->dev, "spi-max-frequency", + &hs->max_freq); + if (ret) { + dev_err(&pdev->dev, + "failed to get max SPI clocking speed, ret=%d\n", ret); + return -EINVAL; + } + + version = readl(hs->regs + HISI_SPI_VERSION); + + ret = hisi_spi_add_host(&pdev->dev, hs); + if (ret) + return ret; + + platform_set_drvdata(pdev, hs); + + dev_info(&pdev->dev, "hw version:0x%x max-freq:%u kHz\n", + version, hs->max_freq / 1000); + + return 0; +} + +static const struct acpi_device_id hisi_spi_acpi_match[] = { + {"HISI03E1", 0}, + {} +}; +MODULE_DEVICE_TABLE(acpi, hisi_spi_acpi_match); + +static struct platform_driver hisi_spi_driver = { + .probe = hisi_spi_probe, + .driver = { + .name = "hisi_spi", + .acpi_match_table = hisi_spi_acpi_match, + }, +}; +module_platform_driver(hisi_spi_driver); + +MODULE_AUTHOR("Jay Fang <f.fangjian@huawei.com>"); +MODULE_DESCRIPTION("HiSilicon SPI Controller Driver"); +MODULE_LICENSE("GPL v2");
This driver supports SPI Controller for HiSilicon Kunpeng SOCs. This driver supports SPI operations using FIFO mode of transfer. DMA is not supported, and we just use IRQ mode for operation completion notification. Only ACPI firmware is supported. Signed-off-by: Jay Fang <f.fangjian@huawei.com> --- MAINTAINERS | 7 + drivers/spi/Kconfig | 10 + drivers/spi/Makefile | 1 + drivers/spi/spi-hisi.c | 573 +++++++++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 591 insertions(+) create mode 100644 drivers/spi/spi-hisi.c