: What should we do in txd->tx_submit() and device_issue_pending()
Hi Dan/Vinod,
I have one query regarding implementation of few routines in DMA drivers.
What the exact purpose of txd->tx_submit() and device_issue_pending() routines?
Should tx_submit() always start physical transfer (Obviously, if channel is free)? Or should it just queue requests and then finally a call to device_issue_pending() must start physical transfers?
Signature · 1 line
viresh
(pls wrap your email to 80lines, I have reflowed below for readability)
Quoted 9 lines
Hi Dan/Vinod,
I have one query regarding implementation of few routines in DMA drivers.
What the exact purpose of
txd->tx_submit()anddevice_issue_pending()routines?Should
tx_submit()always start physical transfer (Obviously, if channel is free)? Or should it just queue requests and then finally a call todevice_issue_pending()must start physical transfers?
(reflowed your email for 80lines)tx_submit is not supposed to start the transfer, it is supposed to add the descriptors in dma driver's queue. The issue_pending is supposed to actually start the transfer. The reasons were well articulated is discussion we had when we did the DMA documentation.
Pls see https://lkml.org/lkml/2011/5/24/530
~Vinod
Quoted 11 lines · Vinod
(pls wrap your email to 80lines, I have reflowed below for readability)
Quoted 9 lines
Hi Dan/Vinod,
I have one query regarding implementation of few routines in DMA drivers.
What the exact purpose of
txd->tx_submit()anddevice_issue_pending()routines?Should
tx_submit()always start physical transfer (Obviously, if channel is free)? Or should it just queue requests and then finally a call todevice_issue_pending()must start physical transfers?(reflowed your email for 80lines)
Sorry!! :(
Quoted 6 lines
tx_submitis not supposed to start the transfer, it is supposed to add the descriptors in dma driver's queue. Theissue_pendingis supposed to actually start the transfer. The reasons were well articulated is discussion we had when we did the DMA documentation.
Pls see https://lkml.org/lkml/2011/5/24/530
Thanks.
Signature · 1 line
viresh