Re: [linus:master] [eventfs] 852e46e239: BUG:unable_to_handle_page_fault_for_address

[Linus Torvalds]

On Tue, 30 Jan 2024 at 10:23, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
> I know you don't send patches inlined anymore, which is a shame, because
> patchwork takes care of all the administering when patches are inlined, and
> I don't miss patches like I use to.

I just sent the whole series as individual patches., and doing a

  b4 am 20240130190355.11486-1-torvalds@xxxxxxxxxxxxxxxxxxxx

should get them all (or if you use patchwork, do that)

I don't normally do these inlined patches any more, because honestly,
99% of all patches I do end up being "I can't test this very well, I
think you should do something like this".

In fact, for simple ones where I might not have even compile-tested
them, much less booted into a kernel with them, I will actively
whitespace-corrupt the patch, just to make sure they aren't applied as
any kind of real workflow - they are almost always meant as a "I think
you should do this, and take credit for it all".

And so when I'm working on a series like this, I'll send attachments
just because it's easier, and because I don't want to patch-bomb
people with some kind of crazy work-in-progress thing.

But I'm reasonably comfortable with this series now, so I sent it as a
"real" patch series. I like it partly because it just removes a lot of
lines:

 3 files changed, 160 insertions(+), 433 deletions(-)

but mostly because the lines it removes are what I consider actively
broken code. So it's not just getting rid of LOC, it's getting rid of
complexity (and bugs) IMHO.

That said, I also don't think that patch series is any kind of
"final". I didn't fix up the readdir iterator locking, for example. I
don't think the SRCU parts are needed at all any more thanks to the
refcounting - the 'ei' is no longer protected by SRCU, it's protected
by virtue of us having the file open (and thus holding the dentry).

So please think of that series not as any kind of final word. More as
a "I strongly believe this is the direction eventfs should go".

I am perfectly ok with you munging those patches and taking full
credit for them, for example.

My "testing" has not involved any real tracing usage, and while I
*have* booted that thing, and have done some very basic smoke-testing
in /sys/kernel/tracing, 99% of the work was literally me just going
through the lookup code, removing everything I found objectionable,
and replacing it with what the VFS layer generally would want.

                      Linus