On Fri, 04 Jun 2010 07:59:42 -0700, "Venkateswararao Jujjuri (JV)" <jv...@linux.vnet.ibm.com> wrote: > Aneesh Kumar K. V wrote: > > On Thu, 3 Jun 2010 18:29:02 +0530, Sripathi Kodi <sripat...@in.ibm.com> > > wrote: > >> On Wed, 02 Jun 2010 19:49:24 +0530 > >> "Aneesh Kumar K. V" <aneesh.ku...@linux.vnet.ibm.com> wrote: > >> > >>> On Fri, 28 May 2010 16:08:43 +0530, Sripathi Kodi <sripat...@in.ibm.com> > >>> wrote: > >>>> From: M. Mohan Kumar <mo...@in.ibm.com> > >>>> > >>>> SYNOPSIS > >>>> > >>>> size[4] Tgetattr tag[2] fid[4] > >>>> > >>>> size[4] Rgetattr tag[2] lstat[n] > >>>> > >>>> DESCRIPTION > >>>> > >>>> The getattr transaction inquires about the file identified by fid. > >>>> The reply will contain a machine-independent directory entry, > >>>> laid out as follows: > >>>> > >>>> qid.type[1] > >>>> the type of the file (directory, etc.), represented as a bit > >>>> vector corresponding to the high 8 bits of the file's mode > >>>> word. > >>>> > >>>> qid.vers[4] > >>>> version number for given path > >>>> > >>>> qid.path[8] > >>>> the file server's unique identification for the file > >>>> > >>>> st_mode[4] > >>>> Permission and flags > >>>> > >>>> st_nlink[8] > >>>> Number of hard links > >>>> > >>>> st_uid[4] > >>>> User id of owner > >>>> > >>>> st_gid[4] > >>>> Group ID of owner > >>>> > >>>> st_rdev[8] > >>>> Device ID (if special file) > >>>> > >>>> st_size[8] > >>>> Size, in bytes > >>>> > >>>> st_blksize[8] > >>>> Block size for file system IO > >>> > >>> So it should be scaled by iounit right ? If we say 9p block size is > >>> iounit. > >> Yes, I think it should be iounit. Currently st_blksize being returned > >> in stat structure to the user space does not use this field that comes > >> from the server. It is being calculated as follows in > >> generic_fillattr(): > >> > >> stat->blksize = (1 << inode->i_blkbits); > >> > >> So there may not be a need to put st_blksize on the protocol. Further, > >> inode->i_blkbits is copied from sb->s_blocksize_bits. For 9P this value > >> is obtained as: > > > > That is what linux kernel currently does. But from the protocol point of > > view and not looking at specific linux implementation i would suggest to > > put st_blksize on wire. > > This is part of .L protocol. Specifically for Linux systems. So, if Linux is > already > doing it, I don't think we need to repeat it. >
But nothing prevents from changing Linux internal implementation. So we can't depend on Linux kernel internal implementation. Later in 2.6.x we may not derive stat->blksize from inode->i_blkbits at all. So we cannot depend on Linux kernel internal implementation. -aneesh