Re: [PATCH] gfp: Include __GFP_NOWARN in GFP_NOWAIT

[Andrew Morton]

On Thu,  9 Nov 2023 21:15:07 +0000 "Matthew Wilcox (Oracle)" <willy@xxxxxxxxxxxxx> wrote:

> GFP_NOWAIT callers are always prepared for their allocations to fail
> because they fail so frequently.  Forcing the callers to remember to add
> __GFP_NOWARN is just annoying and leads to an endless stream of patches
> for the places where we forgot to add it.

A possible problem is call sites where the developers were relying upon
the core MM's warning, so they chose to omit __GFP_NOWARN rather than
adding a local printk.

We're now silencing core MM's warning so those developers might lose
some diagnostic information.

Random example, arch/s390/pci/pci_clp.c:clp_refresh_fh().  After this
change, the pci_clp developers won't see (or be told about) allocation
failures in this function.  If they know about this change, they may
now choose to add a local printk.

It's not the end of the world by any means.  One possible way to
prevent this change in behavior is to add a new __GFP_WARN and go add
it to all the sites which use bare __GFP_NOWAIT and which do not have a
local printk for the allocation failure.