On 20 Nov 2024, at 2:57, Cedric Blancher wrote: > On Tue, 19 Nov 2024 at 18:07, Steve Dickson <steved@xxxxxxxxxx> wrote: >> >> From: Seiichi Ikarashi <s.ikarashi@xxxxxxxxxxx> >> >> The rsize/wsize values are not multiples of 1024 but multiples of the >> system's page size or powers of 2 if < system's page size as defined >> in fs/nfs/internal.h:nfs_io_size(). >> >> Signed-off-by: Steve Dickson <steved@xxxxxxxxxx> > > REJECT. As discussed, this is the WRONG approach. I am astounded that this response keeps appearing, we must have some fundamental inability to understand each other. > The pagesize is not not easily determinable (/bin/pagesize not even being > part of the default install), and the page size is flexible on many > architectures. rsize/wsize depending on the page size makes this option > non portable across platforms, or even same platforms with different > default pagesize settings. Why do you need to determine the page size to set this option? You are incorrect when stating that rsize/wsize depends on the page size. As I explained before, one does not need to know the page size to set this mount option. The kernel takes the desired size and rounds down to the current system's page size. > In real life, this can prevent puppet from working for NFS root, if > NFS root needs rsize/wsize, and someone switches the default page > size. What does "prevent puppet from working" mean? Please show a real example of this problem that requires a kernel fix. > I thought the correct fix would be to fix the NFS client to count in > kbytes as documented, and round up/down to the pagesize. What problem are we fixing? Currently, the only non-hypothetical issue is that the man page is incorrect. Ben
