intel_mid_dma: change the slave interface
PATCH 6/6 intel_mid_dma: change the slave interface
In 2.6.36 kernel, dma slave control command was introduced, this patch changes the intel-mid-dma driver to this new kernel slave interface
--- drivers/dma/intel_mid_dma.c | 66 ++++++++++++++++++++++++------------- drivers/dma/intel_mid_dma_regs.h | 11 +++++- include/linux/intel_mid_dma.h | 15 +-------- 3 files changed, 53 insertions(+), 39 deletions(-)
drivers/dma/intel_mid_dma.c +43 -23
@@ -92,13 +92,13 @@ static int get_block_ts(int len, int tx_width, int block_size)int byte_width =0 , block_ts =0 ;switch (tx_width) { -case LNW_DMA_WIDTH_8BIT : +case DMA_SLAVE_BUSWIDTH_1_BYTE : byte_width =1 ;break ; -case LNW_DMA_WIDTH_16BIT : +case DMA_SLAVE_BUSWIDTH_2_BYTES : byte_width =2 ;break ; -case LNW_DMA_WIDTH_32BIT : +case DMA_SLAVE_BUSWIDTH_4_BYTES :default : byte_width =4 ;break ;
@@ -367,7 +367,7 @@ static int midc_lli_fill_sg(struct intel_mid_dma_chan *midc,int i;pr_debug ("MDMA: Entered midc_lli_fill_sg\n" ); - mids = midc->chan.private; + mids = midc->mid_slave; lli_bloc_desc = desc->lli; lli_next = desc->lli_phys;
@@ -398,9 +398,9 @@ static int midc_lli_fill_sg(struct intel_mid_dma_chan *midc, sg_phy_addr =sg_phys (sg);if (desc->dirn ==DMA_TO_DEVICE ) { lli_bloc_desc->sar = sg_phy_addr; - lli_bloc_desc->dar = mids->per_addr; + lli_bloc_desc->dar = mids->dma_slave.dst_addr; }else if (desc->dirn ==DMA_FROM_DEVICE ) { - lli_bloc_desc->sar = mids->per_addr; + lli_bloc_desc->sar = mids->dma_slave.src_addr; lli_bloc_desc->dar = sg_phy_addr; }/*Copy values into block descriptor in system memroy*/
@@ -507,6 +507,23 @@ static enum dma_status intel_mid_dma_tx_status(struct dma_chan *chan,return ret; } +static int dma_slave_control (struct dma_chan *chan,unsigned long arg) +{ +struct intel_mid_dma_chan *midc =to_intel_mid_dma_chan (chan); +struct dma_slave_config *slave = (struct dma_slave_config *)arg; +struct intel_mid_dma_slave *mid_slave; + +BUG_ON (!midc); +BUG_ON (!slave); +pr_debug ("MDMA: slave control called\n" ); + + mid_slave =to_intel_mid_dma_slave (slave); + +BUG_ON (!mid_slave); + + midc->mid_slave = mid_slave; +return 0 ; +}/** * intel_mid_dma_device_control - DMA device control * @chan: chan for DMA control
@@ -523,6 +540,9 @@ static int intel_mid_dma_device_control(struct dma_chan *chan,struct intel_mid_dma_desc *desc, *_desc;union intel_mid_dma_cfg_lo cfg_lo; +if (cmd ==DMA_SLAVE_CONFIG ) +return dma_slave_control (chan, arg); +if (cmd !=DMA_TERMINATE_ALL )return -ENXIO ;
@@ -540,7 +560,6 @@ static int intel_mid_dma_device_control(struct dma_chan *chan,/* Disable interrupts */ disable_dma_interrupt (midc); midc->descs_allocated =0 ; - midc->slave =NULL ;spin_unlock_bh (&midc->lock);list_for_each_entry_safe (desc, _desc, &midc->active_list, desc_node) {
@@ -578,23 +597,24 @@ static struct dma_async_tx_descriptor *intel_mid_dma_prep_memcpy(union intel_mid_dma_ctl_hi ctl_hi;union intel_mid_dma_cfg_lo cfg_lo;union intel_mid_dma_cfg_hi cfg_hi; -enum intel_mid_dma_width width =0 ; +enum dma_slave_buswidth width;pr_debug ("MDMA: Prep for memcpy\n" );BUG_ON (!chan);if (!len)return NULL ; - mids = chan->private; -BUG_ON (!mids); - midc =to_intel_mid_dma_chan (chan);BUG_ON (!midc); + mids = midc->mid_slave; +BUG_ON (!mids); +pr_debug ("MDMA:called for DMA %x CH %d Length %zu\n" , midc->dma->pci_id, midc->ch_id, len);pr_debug ("MDMA:Cfg passed Mode %x, Dirn %x, HS %x, Width %x\n" , - mids->cfg_mode, mids->dirn, mids->hs_mode, mids->src_width); + mids->cfg_mode, mids->dma_slave.direction, + mids->hs_mode, mids->dma_slave.src_addr_width);/*calculate CFG_LO*/ if (mids->hs_mode ==LNW_DMA_SW_HS ) {
@@ -613,13 +633,13 @@ static struct dma_async_tx_descriptor *intel_mid_dma_prep_memcpy(if (midc->dma->pimr_mask) { cfg_hi.cfgx.protctl =0x0 ;/*default value*/ cfg_hi.cfgx.fifo_mode =1 ; -if (mids->dirn ==DMA_TO_DEVICE ) { +if (mids->dma_slave.direction ==DMA_TO_DEVICE ) { cfg_hi.cfgx.src_per =0 ;if (mids->device_instance ==0 ) cfg_hi.cfgx.dst_per =3 ;if (mids->device_instance ==1 ) cfg_hi.cfgx.dst_per =1 ; - }else if (mids->dirn ==DMA_FROM_DEVICE ) { + }else if (mids->dma_slave.direction ==DMA_FROM_DEVICE ) {if (mids->device_instance ==0 ) cfg_hi.cfgx.src_per =2 ;if (mids->device_instance ==1 )
@@ -636,7 +656,7 @@ static struct dma_async_tx_descriptor *intel_mid_dma_prep_memcpy(/*calculate CTL_HI*/ ctl_hi.ctlx.reser =0 ; ctl_hi.ctlx.done =0 ; - width = mids->src_width; + width = mids->dma_slave.src_addr_width; ctl_hi.ctlx.block_ts =get_block_ts (len, width, midc->dma->block_size); pr_debug("MDMA:calc len %d for block size %d\n",
@@ -644,21 +664,21 @@ static struct dma_async_tx_descriptor *intel_mid_dma_prep_memcpy(/*calculate CTL_LO*/ ctl_lo.ctl_lo =0 ; ctl_lo.ctlx.int_en =1 ; - ctl_lo.ctlx.dst_tr_width = mids->dst_width; - ctl_lo.ctlx.src_tr_width = mids->src_width; - ctl_lo.ctlx.dst_msize = mids->src_msize; - ctl_lo.ctlx.src_msize = mids->dst_msize; + ctl_lo.ctlx.dst_tr_width = mids->dma_slave.dst_addr_width; + ctl_lo.ctlx.src_tr_width = mids->dma_slave.src_addr_width; + ctl_lo.ctlx.dst_msize = mids->dma_slave.src_maxburst; + ctl_lo.ctlx.src_msize = mids->dma_slave.dst_maxburst;if (mids->cfg_mode ==LNW_DMA_MEM_TO_MEM ) { ctl_lo.ctlx.tt_fc =0 ; ctl_lo.ctlx.sinc =0 ; ctl_lo.ctlx.dinc =0 ; }else { -if (mids->dirn ==DMA_TO_DEVICE ) { +if (mids->dma_slave.direction ==DMA_TO_DEVICE ) { ctl_lo.ctlx.sinc =0 ; ctl_lo.ctlx.dinc =2 ; ctl_lo.ctlx.tt_fc =1 ; - }else if (mids->dirn ==DMA_FROM_DEVICE ) { + }else if (mids->dma_slave.direction ==DMA_FROM_DEVICE ) { ctl_lo.ctlx.sinc =2 ; ctl_lo.ctlx.dinc =0 ; ctl_lo.ctlx.tt_fc =2 ;
@@ -681,7 +701,7 @@ static struct dma_async_tx_descriptor *intel_mid_dma_prep_memcpy( desc->ctl_lo = ctl_lo.ctl_lo; desc->ctl_hi = ctl_hi.ctl_hi; desc->width = width; - desc->dirn = mids->dirn; + desc->dirn = mids->dma_slave.direction; desc->lli_phys =0 ; desc->lli =NULL ; desc->lli_pool =NULL ;
@@ -722,7 +742,7 @@ static struct dma_async_tx_descriptor *intel_mid_dma_prep_slave_sg( midc =to_intel_mid_dma_chan (chan);BUG_ON (!midc); - mids = chan->private; + mids = midc->mid_slave;BUG_ON (!mids);if (!midc->dma->pimr_mask) {
drivers/dma/intel_mid_dma_regs.h +9 -2
@@ -187,13 +187,13 @@ struct intel_mid_dma_chan {struct list_head active_list;struct list_head queue;struct list_head free_list; -struct intel_mid_dma_slave *slave;unsigned int descs_allocated;struct middma_device *dma;bool busy;bool in_use;u32 raw_tfr;u32 raw_block; +struct intel_mid_dma_slave *mid_slave; }; static inline struct intel_mid_dma_chan *to_intel_mid_dma_chan(
@@ -264,7 +264,7 @@ struct intel_mid_dma_desc {dma_addr_t next;enum dma_data_direction dirn;enum dma_status status; -enum intel_mid_dma_width width;/*width of DMA txn*/ +enum dma_slave_buswidth width;/*width of DMA txn*/ enum intel_mid_dma_mode cfg_mode;/*mode configuration*/ };
@@ -289,6 +289,13 @@ static inline struct intel_mid_dma_desc *to_intel_mid_dma_desc return container_of(txd, struct intel_mid_dma_desc, txd); } +static inline struct intel_mid_dma_slave *to_intel_mid_dma_slave + (struct dma_slave_config *slave) +{ + return container_of(slave, struct intel_mid_dma_slave, dma_slave); +} + +int dma_resume (struct pci_dev *pci);#endif /*__INTEL_MID_DMAC_REGS_H__*/
include/linux/intel_mid_dma.h +1 -14
@@ -28,14 +28,6 @@#include <linux/dmaengine.h> #define DMA_PREP_CIRCULAR_LIST (1 << 10) -/*DMA transaction width, src and dstn width would be same -The DMA length must be width aligned, -for 32 bit width the length must be 32 bit (4bytes) aligned only*/ -enum intel_mid_dma_width { -LNW_DMA_WIDTH_8BIT =0x0 , -LNW_DMA_WIDTH_16BIT =0x1 , -LNW_DMA_WIDTH_32BIT =0x2 , -};/*DMA mode configurations*/ enum intel_mid_dma_mode {
@@ -75,15 +67,10 @@ enum intel_mid_dma_msize {* peripheral device connected to single DMAC */ struct intel_mid_dma_slave { -enum dma_data_direction dirn; -enum intel_mid_dma_width src_width;/*width of DMA src txn*/ -enum intel_mid_dma_width dst_width;/*width of DMA dst txn*/ enum intel_mid_dma_hs_mode hs_mode;/*handshaking*/ enum intel_mid_dma_mode cfg_mode;/*mode configuration*/ -enum intel_mid_dma_msize src_msize;/*size if src burst*/ -enum intel_mid_dma_msize dst_msize;/*size of dst burst*/ -dma_addr_t per_addr;/*Peripheral address*/ unsigned int device_instance;/*0, 1 for periphral instance*/ +struct dma_slave_config dma_slave; };#endif /*__INTEL_MID_DMA_H__*/
Signature · 1 line
1.6.6.1
Quoted 6 lines · Vinod Koul
In 2.6.36 kernel, dma slave control command was introduced, this patch changes the intel-mid-dma driver to this new kernel slave interface
Signed-off-by Vinod Koul <[email protected]>---
[..]
drivers/dma/intel_mid_dma.c static enum dma_status intel_mid_dma_tx_status(struct dma_chan *chan, · line 520
Quoted 11 lines
drivers/dma/intel_mid_dma.c +11
-0+static int dma_slave_control (struct dma_chan *chan,unsigned long arg) +{ + struct intel_mid_dma_chan *midc = to_intel_mid_dma_chan(chan); + struct dma_slave_config *slave = (struct dma_slave_config *)arg; + struct intel_mid_dma_slave *mid_slave; + + BUG_ON(!midc); + BUG_ON(!slave); + pr_debug("MDMA: slave control called\n"); + + mid_slave = to_intel_mid_dma_slave(slave);
I know this was the initial recommendation to support custom extensions to dma_slave_config, but this isn't safe. How does this routine know that that 'slave' is embedded in an intel_mid_dma_slave structure. I applied it as is, but I think we need a better approach.
Maybe a per device alloc_dma_slave_config() such that ->dma_control() can assume the driver allocated the dma_slave_config. As it stands nothing in the calling convention guarantees type safety.
--
Dan
Quoted 25 lines · On Fri, Oct 8, 2010 at 10:58 PM, [email protected] wrote:
Quoted 6 lines · Vinod Koul
In 2.6.36 kernel, dma slave control command was introduced, this patch changes the intel-mid-dma driver to this new kernel slave interface
Signed-off-by Vinod Koul <[email protected]>---
[..]
drivers/dma/intel_mid_dma.c static enum dma_status intel_mid_dma_tx_status(struct dma_chan *chan, · line 520
Quoted 11 lines
drivers/dma/intel_mid_dma.c +11
-0+static int dma_slave_control (struct dma_chan *chan,unsigned long arg) +{ + struct intel_mid_dma_chan *midc = to_intel_mid_dma_chan(chan); + struct dma_slave_config *slave = (struct dma_slave_config *)arg; + struct intel_mid_dma_slave *mid_slave; + + BUG_ON(!midc); + BUG_ON(!slave); + pr_debug("MDMA: slave control called\n"); + + mid_slave = to_intel_mid_dma_slave(slave);I know this was the initial recommendation to support custom extensions to
dma_slave_config, but this isn't safe. How does this routine know that that 'slave' is embedded in anintel_mid_dma_slavestructure. I applied it as is, but I think we need a better approach.
Quoted 3 lines
Maybe a per device
alloc_dma_slave_config()such that ->dma_control() can assume the driver allocated thedma_slave_config. As it stands nothing in the calling convention guarantees type safety.
Dan,
Thanks for applying the patches.
I had thought thru this earlier, yes I agree that it is not fully safe to *assume* that intel_mid_dma_slave embeds the dma_slave_config structure.
But we can also argue that since this is DMA_SLAVE device and device needs to always pass controller specific arguments and hence knows about this controller, so will have such a thing in place.
Yes but this can be easily abused.
Another alternative is to maintain the channel->private pointer for these sort of details as we were doing previously...
Thoughts...?
~Vinod