s3cmci SDIO patches
cover s3cmci SDIO patches
Hi Ben,
here are the s3cmci SDIO patches that I've done so far.
The first patch modifies the s3cmci driver so that it supports non-word-sized transfers which is needed by SDIO.
The second patch removes calls the pio_tasklet directly from the interrupt handler instead of scheduling it as a tasklet. This reduces the CPU load for the driver sligthtly.
The third patch adds experimental SDIO interrupt support. Since there are issues with spurious interrupts on some S3C processors, I think this needs more testing before going into the kernel.
I'm working on trying to get the SDIO driver to go faster and use less CPU, so I'm mainly doing two things right now, first of all I'd like to try to make DMA work again. There are hooks in the code for DMA, but when i tried turning it on, it didn't work, so I'll try to chase that down.
I'll probably make some other random changes, I'd like to make the driver support non-word-aligned block transfers for example, something that it doesn't support right now. It doesn't seem to be needed for SDIO Bluetooth Type A, but it would be nice to have it for correctness.
And I'll try to clean up my patches to add asynchronous operations to the SDIO layer, which might cause some fallout on the s3cmci driver.
Is there anything you'd like me to do with the s3cmci driver while I'm at it?
/Christer
Signature · 3 lines
"Just how much can I get away with and still go to heaven?" Christer Weinigel <[email protected]> http://www.weinigel.se
PATCH 1/3 s3cmci -- support odd block transfers
To be able to do SDIO the s3cmci driver has to support non-word-sized transfers. Change pio_words into pio_bytes and fix up all the places where it is used.
This variant of the patch will not overrun the buffer when reading an odd number of bytes. When writing, this variant will still read past the end of the buffer, but since the driver can't support non-word- aligned transfers anyway, this should not be a problem, since a word-aligned transfer will never cross a page boundary.
This has been tested with a CSR SDIO Bluetooth Type A device on a Samsung S3C24A0 processor.
drivers/mmc/host/s3cmci.c +53 -27
@@ -191,7 +191,7 @@ static inline void clear_imask(struct s3 }static inline int get_data_buffer (struct s3cmci_host *host, -u32 *words,u32 **pointer) +u32 *bytes,u32 **pointer) {struct scatterlist *sg;
@@ -208,7 +208,7 @@ static inline int get_data_buffer(struct } sg = &host->mrq->data->sg[host->pio_sgptr]; - *words = sg->length >>2 ; + *bytes = sg->length; *pointer =sg_virt (sg); host->pio_sgptr++;
@@ -224,7 +224,7 @@ static inline u32 fifo_count(struct s3cmu32 fifostat =readl (host->base +S3C2410_SDIFSTA ); fifostat &=S3C2410_SDIFSTA_COUNTMASK ; -return fifostat >>2 ; +return fifostat; }static inline u32 fifo_free (struct s3cmci_host *host)
@@ -232,13 +232,14 @@ static inline u32 fifo_free(struct s3cmcu32 fifostat =readl (host->base +S3C2410_SDIFSTA ); fifostat &=S3C2410_SDIFSTA_COUNTMASK ; -return (63 - fifostat) >>2 ; +return 63 - fifostat; }static void do_pio_read (struct s3cmci_host *host) {int res;u32 fifo; +u32 fifo_words;void __iomem *from_ptr;/* write real prescaler to host, it might be set slow to fix */
@@ -247,8 +248,8 @@ static void do_pio_read(struct s3cmci_ho from_ptr = host->base + host->sdidata;while ((fifo =fifo_count (host))) { -if (!host->pio_words) { - res =get_data_buffer (host, &host->pio_words, +if (!host->pio_bytes) { + res =get_data_buffer (host, &host->pio_bytes, &host->pio_ptr);if (res) { host->pio_active =XFER_NONE ;
@@ -261,26 +262,45 @@ static void do_pio_read(struct s3cmci_hodbg (host, dbg_pio,"pio_read(): new target: [%i]@[%p]\n" , - host->pio_words, host->pio_ptr); + host->pio_bytes, host->pio_ptr); }dbg (host, dbg_pio,"pio_read(): fifo:[%02i] buffer:[%03i] dcnt:[%08X]\n" , - fifo, host->pio_words, + fifo, host->pio_bytes,readl (host->base +S3C2410_SDIDCNT )); -if (fifo > host->pio_words) - fifo = host->pio_words; +/* If we have reached the end of the block, we can +* read a word and get 1 to 3 bytes. If we in the +* middle of the block, we have to read full words, +* otherwise we will write garbage, so round down to +* an even multiple of 4. */ +if (fifo >= host->pio_bytes) + fifo = host->pio_bytes; +else + fifo -= fifo &3 ; - host->pio_words -= fifo; + host->pio_bytes -= fifo; host->pio_count += fifo; -while (fifo--) + fifo_words = fifo >>2 ; +while (fifo_words--) *(host->pio_ptr++) =readl (from_ptr); + +if (fifo &3 ) { +u32 n = fifo &3 ; +u32 data =readl (from_ptr); +u8 *p = (u8 *)host->pio_ptr; + +while (n--) { + *p++ = data; + data >>=8 ; + } + } } -if (!host->pio_words) { - res =get_data_buffer (host, &host->pio_words, &host->pio_ptr); +if (!host->pio_bytes) { + res =get_data_buffer (host, &host->pio_bytes, &host->pio_ptr);if (res) {dbg (host, dbg_pio,"pio_read(): complete (no more buffers).\n" );
@@ -304,8 +324,8 @@ static void do_pio_write(struct s3cmci_h to_ptr = host->base + host->sdidata;while ((fifo =fifo_free (host))) { -if (!host->pio_words) { - res =get_data_buffer (host, &host->pio_words, +if (!host->pio_bytes) { + res =get_data_buffer (host, &host->pio_bytes, &host->pio_ptr);if (res) { dbg(host, dbg_pio,
@@ -317,16 +337,23 @@ static void do_pio_write(struct s3cmci_hdbg (host, dbg_pio,"pio_write(): new source: [%i]@[%p]\n" , - host->pio_words, host->pio_ptr); + host->pio_bytes, host->pio_ptr); } -if (fifo > host->pio_words) - fifo = host->pio_words; +/* If we have reached the end of the block, we have to +* write exactly the remaining number of bytes. If we +* in the middle of the block, we have to write full +* words, so round down to an even multiple of 4. */ +if (fifo >= host->pio_bytes) + fifo = host->pio_bytes; +else + fifo -= fifo &3 ; - host->pio_words -= fifo; + host->pio_bytes -= fifo; host->pio_count += fifo; + fifo = (fifo +3 ) >>2 ;while (fifo--)writel (*(host->pio_ptr++), to_ptr); }
@@ -351,9 +378,9 @@ static void pio_tasklet(unsigned long daclear_imask (host);if (host->pio_active !=XFER_NONE ) {dbg (host, dbg_err,"unfinished %s " -"- pio_count:[%u] pio_words:[%u]\n" , +"- pio_count:[%u] pio_bytes:[%u]\n" , (host->pio_active ==XFER_READ ) ?"read" :"write" , - host->pio_count, host->pio_words); + host->pio_count, host->pio_bytes);if (host->mrq->data) host->mrq->data->error = -EINVAL ;
@@ -813,11 +840,10 @@ static int s3cmci_setup_data(struct s3cm/* We cannot deal with unaligned blocks with more than * one block being transfered. */ -if (data->blocks >1 ) +if (data->blocks >1 ) { +pr_warning ("%s: can't do non-word sized block transfers (blksz %d)\n" , __func__, data->blksz);return -EINVAL ; - -/* No support yet for non-word block transfers. */ -return -EINVAL ; + } } while (readl(host->base + S3C2410_SDIDSTA) &
@@ -897,7 +923,7 @@ static int s3cmci_prepare_pio(struct s3cBUG_ON ((data->flags &BOTH_DIR ) ==BOTH_DIR ); host->pio_sgptr =0 ; - host->pio_words =0 ; + host->pio_bytes =0 ; host->pio_count =0 ; host->pio_active = rw ?XFER_WRITE :XFER_READ ;
drivers/mmc/host/s3cmci.h +1 -1
@@ -51,7 +51,7 @@ struct s3cmci_host {int dma_complete;u32 pio_sgptr; -u32 pio_words; +u32 pio_bytes;u32 pio_count;u32 *pio_ptr;#define XFER_NONE 0
Signature · 3 lines
"Just how much can I get away with and still go to heaven?" Christer Weinigel <[email protected]> http://www.weinigel.se
PATCH 2/3 s3cmci - call pio_tasklet from IRQ
In reply to Christer WeinigelScheduling a tasklet to perform the pio transfer introduces a bit of extra processing, just call pio_tasklet directly from the interrupt instead. Writing up to 64 bytes to a FIFO is probably uses less CPU than scheduling a tasklet anyway.
drivers/mmc/host/s3cmci.c +7 -11
@@ -361,11 +361,8 @@ static void do_pio_write(struct s3cmci_henable_imask (host,S3C2410_SDIIMSK_TXFIFOHALF ); } -static void pio_tasklet (unsigned long data) +static void pio_tasklet (struct s3cmci_host *host) { -struct s3cmci_host *host = (struct s3cmci_host *) data; - -disable_irq (host->irq);if (host->pio_active ==XFER_WRITE )
@@ -460,10 +457,10 @@ static irqreturn_t s3cmci_irq(int irq, vif (!host->dodma) {if ((host->pio_active ==XFER_WRITE ) && (mci_fsta &S3C2410_SDIFSTA_TFDET )) { -disable_imask (host,S3C2410_SDIIMSK_TXFIFOHALF ); -tasklet_schedule (&host->pio_tasklet); +pio_tasklet (host); host->status ="pio tx" ; + goto irq_out; } if ((host->pio_active == XFER_READ) &&
@@ -473,8 +470,9 @@ static irqreturn_t s3cmci_irq(int irq, vS3C2410_SDIIMSK_RXFIFOHALF | S3C2410_SDIIMSK_RXFIFOLAST); -tasklet_schedule (&host->pio_tasklet); +pio_tasklet (host); host->status ="pio rx" ; + goto irq_out; } }
@@ -595,7 +593,7 @@ close_transfer: host->complete_what =COMPLETION_FINALIZE ;clear_imask (host); -tasklet_schedule (&host->pio_tasklet); +pio_tasklet (host); goto irq_out;
@@ -666,7 +664,7 @@ void s3cmci_dma_done_callback(struct s3c host->complete_what =COMPLETION_FINALIZE ; out: -tasklet_schedule (&host->pio_tasklet); +pio_tasklet (host);spin_unlock_irqrestore (&host->complete_lock, iflags);return ;
@@ -1198,7 +1196,6 @@ static int __devinit s3cmci_probe(struct }spin_lock_init (&host->complete_lock); -tasklet_init (&host->pio_tasklet, pio_tasklet, (unsigned long ) host);if (is2440) { host->sdiimsk =S3C2440_SDIIMSK ;
@@ -1381,7 +1378,6 @@ static int __devexit s3cmci_remove(strucclk_put (host->clk); -tasklet_disable (&host->pio_tasklet);s3c2410_dma_free (S3CMCI_DMA , &s3cmci_dma_client);free_irq (host->irq, host);
drivers/mmc/host/s3cmci.h +0 -1
@@ -66,5 +66,4 @@ struct s3cmci_host {char *status;unsigned int ccnt, dcnt; -struct tasklet_struct pio_tasklet; };
Signature · 3 lines
"Just how much can I get away with and still go to heaven?" Christer Weinigel <[email protected]> http://www.weinigel.se
Re: [patch 2/3] s3cmci - call pio_tasklet from IRQ
Quoted 4 lines · Christer Weinigel
Scheduling a tasklet to perform the pio transfer introduces a bit of extra processing, just call
pio_taskletdirectly from the interrupt instead. Writing up to 64 bytes to a FIFO is probably uses less CPU than scheduling a tasklet anyway.
Hmm, i'd be interested to find out how long these are taking... I might try and rig up something to test the time being taken via an SMDK.
If the fifo read/writes are taking significant amounts of time, then the pio tasklet will at least improve the interrupt latencies invloved, as iirc we're currently running the main irq handler in IRQ_DISABLED mode to stop any problems with re-enternancy.... I'll check this and see what is going on.
Quoted 92 lines
Signed-off-by Christer Weinigel <[email protected]>Index:
linux-2.6.26.2/drivers/mmc/host/s3cmci.c
===================================================================drivers/mmc/host/s3cmci.c +7
-11@@ -361,11 +361,8 @@ static void do_pio_write(struct s3cmci_henable_imask (host,S3C2410_SDIIMSK_TXFIFOHALF ); } -static void pio_tasklet (unsigned long data) +static void pio_tasklet (struct s3cmci_host *host) { -struct s3cmci_host *host = (struct s3cmci_host *) data; - -disable_irq (host->irq);if (host->pio_active ==XFER_WRITE )@@ -460,10 +457,10 @@ static irqreturn_t s3cmci_irq(int irq, vif (!host->dodma) {if ((host->pio_active ==XFER_WRITE ) && (mci_fsta &S3C2410_SDIFSTA_TFDET )) { -disable_imask (host,S3C2410_SDIIMSK_TXFIFOHALF ); -tasklet_schedule (&host->pio_tasklet); +pio_tasklet (host); host->status ="pio tx" ; + goto irq_out; } if ((host->pio_active == XFER_READ) &&@@ -473,8 +470,9 @@ static irqreturn_t s3cmci_irq(int irq, vS3C2410_SDIIMSK_RXFIFOHALF | S3C2410_SDIIMSK_RXFIFOLAST); -tasklet_schedule (&host->pio_tasklet); +pio_tasklet (host); host->status ="pio rx" ; + goto irq_out; } }@@ -595,7 +593,7 @@ close_transfer: host->complete_what =COMPLETION_FINALIZE ;clear_imask (host); -tasklet_schedule (&host->pio_tasklet); +pio_tasklet (host); goto irq_out;@@ -666,7 +664,7 @@ void s3cmci_dma_done_callback(struct s3c host->complete_what =COMPLETION_FINALIZE ; out: -tasklet_schedule (&host->pio_tasklet); +pio_tasklet (host);spin_unlock_irqrestore (&host->complete_lock, iflags);return ;@@ -1198,7 +1196,6 @@ static int __devinit s3cmci_probe(struct }spin_lock_init (&host->complete_lock); -tasklet_init (&host->pio_tasklet, pio_tasklet, (unsigned long ) host);if (is2440) { host->sdiimsk =S3C2440_SDIIMSK ;@@ -1381,7 +1378,6 @@ static int __devexit s3cmci_remove(strucclk_put (host->clk); -tasklet_disable (&host->pio_tasklet);s3c2410_dma_free (S3CMCI_DMA , &s3cmci_dma_client);free_irq (host->irq, host);Index:
linux-2.6.26.2/drivers/mmc/host/s3cmci.h
===================================================================drivers/mmc/host/s3cmci.h +0
-1@@ -66,5 +66,4 @@ struct s3cmci_host {char *status;unsigned int ccnt, dcnt; -struct tasklet_struct pio_tasklet; };Signature · 4 lines
-- "Just how much can I get away with and still go to heaven?" Christer Weinigel <[email protected]> http://www.weinigel.se
Signature · 5 lines
-- Ben Q: What's a light-year? A: One-third less calories than a regular year.
Re: [patch 2/3] s3cmci - call pio_tasklet from IRQ
Quoted 6 lines · Ben Dooks
Quoted 4 lines · Christer Weinigel
Scheduling a tasklet to perform the pio transfer introduces a bit of extra processing, just call
pio_taskletdirectly from the interrupt instead. Writing up to 64 bytes to a FIFO is probably uses less CPU than scheduling a tasklet anyway.
Quoted 8 lines
Hmm, i'd be interested to find out how long these are taking... I might try and rig up something to test the time being taken via an SMDK.
If the fifo read/writes are taking significant amounts of time, then the pio tasklet will at least improve the interrupt latencies invloved, as iirc we're currently running the main irq handler in
IRQ_DISABLEDmode to stop any problems with re-enternancy.... I'll check this and see what is going on.
It should be possible to set a flag and then call the pio_task after the spin_unlock_irqrestore instead. I didn't want to do that to change as little of the logic as possible, but it's probably better to do that.
I'm also thinking of changing send_request to do a busy wait for commands without data, that will probably need a bit larger changes.
/Christer
Re: [patch 2/3] s3cmci - call pio_tasklet from IRQ
In reply to Ben DooksQuoted 6 lines · Ben Dooks
Quoted 4 lines · Christer Weinigel
Scheduling a tasklet to perform the pio transfer introduces a bit of extra processing, just call
pio_taskletdirectly from the interrupt instead. Writing up to 64 bytes to a FIFO is probably uses less CPU than scheduling a tasklet anyway.
Quoted 8 lines
Hmm, i'd be interested to find out how long these are taking... I might try and rig up something to test the time being taken via an SMDK.
If the fifo read/writes are taking significant amounts of time, then the pio tasklet will at least improve the interrupt latencies invloved, as iirc we're currently running the main irq handler in
IRQ_DISABLEDmode to stop any problems with re-enternancy.... I'll check this and see what is going on.
Ok, I just measured this on the 200 MHz S3C24A0, when running the SDIO bus at 10 MHz, the longest time I saw the driver spend in the pio_read function was ~10us. I guess that means that the hardware managed to empty the fifo enough to do yet another spin through the loop. So with a faster SDIO clock the time spent in pio_read ought to go up, and for a long transfer it could grow without bounds.
I also tried to run the code with the schedule_tasklet still there, and then I saw ~13us as the longest time spent in the loop, and every now and then there was a ~10ms gap when the clock stopped.
If we have working DMA, I think the PIO tasklet is unneccesary, then we'll do PIO for short transfers which won't affect latency much, and use DMA for long transfers that would have affected latency if done with PIO from interrupt context.
/Christer
PATCH 3/3 s3cmci - add SDIO interrupt support
In reply to Christer WeinigelThis adds support for the SDIO irq to the s3cmci driver.
The driver calls to enable_irq and disable_irq are not matched properly when doing SDIO, so I've worked around this by counting the number of enables and disables. This is not the proper way to do it, but makes things work for me the moment on a Samsung S3C24A0 processor. There are supposed to be issues with spurious interrupts on some S3C processors, so this needs more investigation and testing, but I thought this patch may be useful anyway.
drivers/mmc/host/s3cmci.c +50 -9
@@ -187,7 +187,11 @@ static inline u32 disable_imask(struct sstatic inline void clear_imask (struct s3cmci_host *host) { -writel (0 , host->base + host->sdiimsk); +u32 newmask; + + newmask =readl (host->base + host->sdiimsk); + newmask &=S3C2410_SDIIMSK_SDIOIRQ ; +writel (newmask, host->base + host->sdiimsk); } static inline int get_data_buffer(struct s3cmci_host *host,
@@ -364,6 +368,7 @@ static void do_pio_write(struct s3cmci_hstatic void pio_tasklet (struct s3cmci_host *host) {disable_irq (host->irq); + host->irq_disabled++;if (host->pio_active ==XFER_WRITE )do_pio_write (host);
@@ -384,8 +389,12 @@ static void pio_tasklet(struct s3cmci_ho }finalize_request (host); - }else + } + +if (host->irq_disabled) {enable_irq (host->irq); + host->irq_disabled--; + } }/*
@@ -422,6 +431,7 @@ static irqreturn_t s3cmci_irq(int irq, vu32 mci_csta, mci_dsta, mci_fsta, mci_dcnt, mci_imsk;u32 mci_cclear, mci_dclear;unsigned long iflags; +int cardint =0 ;spin_lock_irqsave (&host->complete_lock, iflags);
@@ -433,6 +443,13 @@ static irqreturn_t s3cmci_irq(int irq, v mci_cclear =0 ; mci_dclear =0 ; +if (mci_dsta &S3C2410_SDIDSTA_SDIOIRQDETECT && + mci_imsk &S3C2410_SDIIMSK_SDIOIRQ ) { + cardint =1 ; + mci_dclear =S3C2410_SDIDSTA_SDIOIRQDETECT ; + goto clear_status_bits; + } +if ((host->complete_what ==COMPLETION_NONE ) || (host->complete_what ==COMPLETION_FINALIZE )) { host->status ="nothing to complete" ;
@@ -603,8 +620,12 @@ irq_out: mci_csta, mci_dsta, mci_fsta, mci_dcnt, host->status);spin_unlock_irqrestore (&host->complete_lock, iflags); -return IRQ_HANDLED ; +/* We have to delay this as it calls back into the driver. */ +if (cardint) +mmc_signal_sdio_irq (host->mmc); + +return IRQ_HANDLED ; }/*
@@ -671,7 +692,7 @@ out: fail_request: host->mrq->data->error = -EINVAL ; host->complete_what =COMPLETION_FINALIZE ; -writel (0 , host->base + host->sdiimsk); +clear_imask (host); goto out; }
@@ -716,7 +737,7 @@ static void finalize_request(struct s3cmwritel (0 , host->base +S3C2410_SDICMDARG );writel (S3C2410_SDIDCON_STOP , host->base +S3C2410_SDIDCON );writel (0 , host->base +S3C2410_SDICMDCON ); -writel (0 , host->base + host->sdiimsk); +clear_imask (host);if (cmd->data && cmd->error) cmd->data->error = cmd->error;
@@ -1026,7 +1047,10 @@ static void s3cmci_send_request(struct ms3cmci_send_command (host, cmd);/* Enable Interrupt */ -enable_irq (host->irq); +if (host->irq_disabled) { +enable_irq (host->irq); + host->irq_disabled--; + } }static int s3cmci_card_present (struct s3cmci_host *host)
@@ -1119,9 +1143,9 @@ static void s3cmci_set_ios(struct mmc_ho/* Set CLOCK_ENABLE */ if (ios->clock) - mci_con |=S3C2410_SDICON_CLOCKTYPE ; + mci_con |=S3C2410_SDICON_CLOCKTYPE |S3C2410_SDICON_SDIOIRQ ;else - mci_con &= ~S3C2410_SDICON_CLOCKTYPE ; + mci_con &= ~(S3C2410_SDICON_CLOCKTYPE |S3C2410_SDICON_SDIOIRQ );writel (mci_con, host->base +S3C2410_SDICON );
@@ -1161,10 +1185,26 @@ static int s3cmci_get_ro(struct mmc_hostreturn ret; } +static void s3cmci_enable_sdio_irq (struct mmc_host *mmc,int enable) +{ +struct s3cmci_host *host =mmc_priv (mmc); +unsigned long flags; + +spin_lock_irqsave (&host->complete_lock, flags); + +if (enable) +enable_imask (host,S3C2410_SDIIMSK_SDIOIRQ ); +else +disable_imask (host,S3C2410_SDIIMSK_SDIOIRQ ); + +spin_unlock_irqrestore (&host->complete_lock, flags); +} +static struct mmc_host_ops s3cmci_ops = { .request = s3cmci_request, .set_ios = s3cmci_set_ios, .get_ro = s3cmci_get_ro, + .enable_sdio_irq = s3cmci_enable_sdio_irq, };static struct s3c24xx_mci_pdata s3cmci_def_pdata = {
@@ -1256,6 +1296,7 @@ static int __devinit s3cmci_probe(struct* ensure we don't lock the system with un-serviceable requests. */ disable_irq (host->irq); + host->irq_disabled++; host->irq_cd =s3c2410_gpio_getirq (host->pdata->gpio_detect);
@@ -1301,7 +1342,7 @@ static int __devinit s3cmci_probe(struct mmc->ops = &s3cmci_ops; mmc->ocr_avail =MMC_VDD_32_33 |MMC_VDD_33_34 ; - mmc->caps =MMC_CAP_4_BIT_DATA ; + mmc->caps =MMC_CAP_4_BIT_DATA |MMC_CAP_SDIO_IRQ ; mmc->f_min = host->clk_rate / (host->clk_div *256 ); mmc->f_max = host->clk_rate / host->clk_div;
drivers/mmc/host/s3cmci.h +1 -0
@@ -41,6 +41,7 @@ struct s3cmci_host {unsigned sdidata;int dodma;int dmatogo; +int irq_disabled;struct mmc_request *mrq;int cmd_is_stop;
Signature · 3 lines
"Just how much can I get away with and still go to heaven?" Christer Weinigel <[email protected]> http://www.weinigel.se