Skip to contentSkip to conversation

what happened to byte-addressable read()/write() to character device DAX?

DISCUSSION [email protected] 3 participants 5 messages Oct 19, 2017 – Oct 20, 2017 libnvdimm: non-volatile memory device

lore
Share & tools
lore
#0 lore raw

Hi, Dan,

Referring to

  https://patchwork.kernel.org/project/linux-nvdimm/list/?page=6

  [RFC,2/2] mm, fs: daxfile, an interface for byte-addressable updates to pmem

I'm unable to find the patch in Linux upstream kernel, and am wondering if

there has been a change of plan for providing support to byte granularity

read() and write() to character device /dev/dax.

I've also just started following the email threads, so I might have missed some critical

discussions.  Would you mind to elaborate what's happening to the proposed feature?

Thanks a lot!

Regards,

-jane

_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm

#1 lore raw
Quoted 9 lines · [email protected] writes:

Hi, Dan,

Referring to

https://patchwork.kernel.org/project/linux-nvdimm/list/?page=6

[RFC,2/2] mm, fs: daxfile, an interface for byte-addressable updates to pmem

Quoted 7 lines

I'm unable to find the patch in Linux upstream kernel, and am wondering if there has been a change of plan for providing support to byte granularity read() and write() to character device /dev/dax.

I've also just started following the email threads, so I might have missed some critical discussions. Would you mind to elaborate what's happening to the proposed feature?

Device dax only supports mmap for data path access. There are no plans to support read(2) or write(2).

Cheers,
Jeff
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm

#2 lore raw
Quoted 15 lines · Jeff Moyer
Quoted 13 lines · [email protected] writes:

Hi, Dan,

Referring to

https://patchwork.kernel.org/project/linux-nvdimm/list/?page=6

[RFC,2/2] mm, fs: daxfile, an interface for byte-addressable updates to pmem

I'm unable to find the patch in Linux upstream kernel, and am wondering if there has been a change of plan for providing support to byte granularity read() and write() to character device /dev/dax.

Quoted 6 lines
Quoted 3 lines

I've also just started following the email threads, so I might have missed some critical discussions. Would you mind to elaborate what's happening to the proposed feature?

Device dax only supports mmap for data path access. There are no plans to support read(2) or write(2).

Yes, and the patch set referenced about was not about read(2)/write(2) it was about support for mmap writes without requiring them to be backstopped by calls to msync(). That proposal morphed into the MAP_SYNC and MAP_DIRECT proposals. MAP_SYNC is targeting 4.15. MAP_DIRECT is being reworked into an RDMA-memory-registration-specific interface.
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm

#3 lore raw
Quoted 19 lines · Dan Williams
Quoted 18 lines · Jeff Moyer
Quoted 16 lines · [email protected] writes:

Hi, Dan,

Referring to

https://patchwork.kernel.org/project/linux-nvdimm/list/?page=6

[RFC,2/2] mm, fs: daxfile, an interface for byte-addressable updates to pmem

I'm unable to find the patch in Linux upstream kernel, and am wondering if there has been a change of plan for providing support to byte granularity read() and write() to character device /dev/dax.

I've also just started following the email threads, so I might have missed some critical discussions. Would you mind to elaborate what's happening to the proposed feature?

Quoted 8 lines
Quoted 2 lines

Device dax only supports mmap for data path access. There are no plans to support read(2) or write(2).

Yes, and the patch set referenced about was not about read(2)/write(2) it was about support for mmap writes without requiring them to be backstopped by calls to msync(). That proposal morphed into the MAP_SYNC and MAP_DIRECT proposals. MAP_SYNC is targeting 4.15. MAP_DIRECT is being reworked into an RDMA-memory-registration-specific interface.

Thank you both for the clarification!
Is there documents that elaborate the new resolve in detail?

-jane
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm