SMP TCP userland access

Simon 'corecode' Schubert corecode at fs.ei.tum.de
Thu Jun 7 10:21:18 PDT 2007


Matthew Dillon wrote:
:how does userland access to TCP connections play together with SMP and
:TCP threads?  Is there still a lock?  Would it be beneficial if we lose
:this synchronization and do it more DragonFly-ish?  I.e. will a
:receiving process benefit from a different implementation without locks
:(I've seen sblock() calls, so I guess it is still locked, dunno how this
:plays into the TCP threads thing)
    The sockbuf is access synchronously (aka reading and writing).
    Everything else is accessed asynchronously via a message from the user
    process context to the protocol thread context.  For example,
    listen(), connect(), accept() ,etc... those are actually all messaged.
Do you think it would be beneficial to make the sockbuf access lockless as well?  Beneficial as in:  better throughput for the receiving application (maybe even overall?  imagine the tcp thread being blocked because of readers on other CPUs)

cheers
 simon
--
Serve - BSD     +++  RENT this banner advert  +++    ASCII Ribbon   /"\
Work - Mac      +++  space for low €€€ NOW!1  +++      Campaign     \ /
Party Enjoy Relax   |   http://dragonflybsd.org      Against  HTML   \
Dude 2c 2 the max   !   http://golden-apple.biz       Mail + News   / \
Attachment:
signature.asc
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pgp00003.pgp
Type: application/octet-stream
Size: 252 bytes
Desc: "Description: OpenPGP digital signature"
URL: <http://lists.dragonflybsd.org/pipermail/kernel/attachments/20070607/95331432/attachment-0016.obj>


More information about the Kernel mailing list