Thread (30 messages) 30 messages, 6 authors, 2024-06-06

Re: [PATCH v2 0/4] bugfix: Introduce sendpages_ok() to check sendpage_ok() on contiguous pages

From: Jakub Kicinski <kuba@kernel.org>
Date: 2024-06-04 14:42:26
Also in: ceph-devel, linux-block, linux-nvme

On Tue, 4 Jun 2024 06:30:41 +0200 Christoph Hellwig wrote:
On Sat, Jun 01, 2024 at 03:36:10PM -0700, Jakub Kicinski wrote:
quoted
On Fri, 31 May 2024 09:32:14 +0200 Christoph Hellwig wrote:  
quoted
I still find it hightly annoying that we can't have a helper that
simply does the right thing for callers, but I guess this is the
best thing we can get without a change of mind from the networking
maintainers..  
Change mind about what? Did I miss a discussion?  
Having a net helper to just send some memory in the most efficient
way and leave it up to the networking code to decide if it wants
to use sendpage or sendmsg internally as needed instead of burdening
the caller.
I'd guess the thinking was that if we push back the callers would
switch the relevant allocations to be page-backed. But we can add
a comment above the helper that says "you'd be better off using
page frags and calling sendmsg(MSG_SPLICE_PAGES) directly".
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help