Message-ID: <3187836.aeNJFYEL58@opensuse>
Date:   Fri, 01 Jul 2022 12:10:33 +0200
From:   "Fabio M. De Francesco" <fmdefrancesco@...il.com>
To:     "Eric W. Biederman" <ebiederm@...ssion.com>,
        Ira Weiny <ira.weiny@...el.com>
Cc:     Benjamin LaHaise <bcrl@...ck.org>,
        Alexander Viro <viro@...iv.linux.org.uk>,
        Kees Cook <keescook@...omium.org>,
        Dan Williams <dan.j.williams@...el.com>,
        Matthew Wilcox <willy@...radead.org>, Jan Kara <jack@...e.cz>,
        Jeff Layton <jlayton@...nel.org>,
        Chuck Lever <chuck.lever@...cle.com>,
        Jens Axboe <axboe@...nel.dk>,
        Pavel Begunkov <asml.silence@...il.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Paul Walmsley <paul.walmsley@...ive.com>,
        Palmer Dabbelt <palmer@...belt.com>,
        Albert Ou <aou@...s.berkeley.edu>,
        Nathan Chancellor <nathan@...nel.org>,
        Nick Desaulniers <ndesaulniers@...gle.com>,
        Tom Rix <trix@...hat.com>, linux-aio@...ck.org,
        linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-mm@...ck.org, nvdimm@...ts.linux.dev,
        io-uring@...r.kernel.org, linux-riscv@...ts.infradead.org,
        llvm@...ts.linux.dev
Subject: Re: [PATCH] fs: Replace kmap{,_atomic}() with kmap_local_page()

On giovedì 30 giugno 2022 19:38:08 CEST Eric W. Biederman wrote:
> "Fabio M. De Francesco" <fmdefrancesco@...il.com> writes:
> 
> > The use of kmap() and kmap_atomic() are being deprecated in favor of
> > kmap_local_page().
> >
> > With kmap_local_page(), the mappings are per thread, CPU local and not
> > globally visible. Furthermore, the mappings can be acquired from any
> > context (including interrupts).
> >
> > Therefore, use kmap_local_page() in exec.c because these mappings are 
per
> > thread, CPU local, and not globally visible.
> >
> > Tested with xfstests on a QEMU + KVM 32-bits VM booting a kernel with
> > HIGHMEM64GB enabled.
> 
> Can someone please refresh my memory on what is going on.
> 
> I remember there were limitations that kmap_atomic had that are hard to
> meet so something I think it was kmap_local was invented and created
> to be the kmap_atomic replacement.

Please read highmem.rst. I've updated that document weeks ago:
https://docs.kernel.org/vm/highmem.html?highlight=highmem

Currently it contains many more information I can ever place here in order 
to answer your questions.

Believe me, this is not by any means a way to elude your questions. I'm 
pretty sure that by reading that document you'll have a clear vision on 
what is going on :-)

> 
> What are the requirements on kmap_local?  In copy_strings
> kmap is called in contexts that can sleep in page faults

No problems with kmap_local_page() with regard to page faults (again, 
please read the above-mentioned document).

From that document...

"It’s valid to take pagefaults in a local kmap region []".

"Each call of kmap_atomic() in the kernel creates a non-preemptible section 
and disable pagefaults. This could be a source of unwanted latency. 
Therefore users should prefer kmap_local_page() instead of kmap_atomic().".

> so any
> nearly any requirement except a thread local use is invalidated.
> 
> As you have described kmap_local above it does not sound like kmap_local
> is safe in this context,

Sorry, probably I should add that taking page faults is allowed. Would you 
prefer I send a v2 and add this information?

Thanks,

Fabio

> but that could just be a problem in description
> that my poor memory does is not recalling the necessary details to
> correct.
> 
> Eric
> 
> > Suggested-by: Ira Weiny <ira.weiny@...el.com>
> > Signed-off-by: Fabio M. De Francesco <fmdefrancesco@...il.com>
> > ---
> >  fs/exec.c | 14 +++++++-------
> >  1 file changed, 7 insertions(+), 7 deletions(-)
> >
> > diff --git a/fs/exec.c b/fs/exec.c
> > index 0989fb8472a1..4a2129c0d422 100644
> > --- a/fs/exec.c
> > +++ b/fs/exec.c
> > @@ -583,11 +583,11 @@ static int copy_strings(int argc, struct 
user_arg_ptr argv,
> >  
> >  				if (kmapped_page) {
> >  					
flush_dcache_page(kmapped_page);
> > -					kunmap(kmapped_page);
> > +					kunmap_local(kaddr);
> >  					
put_arg_page(kmapped_page);
> >  				}
> >  				kmapped_page = page;
> > -				kaddr = kmap(kmapped_page);
> > +				kaddr = 
kmap_local_page(kmapped_page);
> >  				kpos = pos & PAGE_MASK;
> >  				flush_arg_page(bprm, kpos, 
kmapped_page);
> >  			}
> > @@ -601,7 +601,7 @@ static int copy_strings(int argc, struct 
user_arg_ptr argv,
> >  out:
> >  	if (kmapped_page) {
> >  		flush_dcache_page(kmapped_page);
> > -		kunmap(kmapped_page);
> > +		kunmap_local(kaddr);
> >  		put_arg_page(kmapped_page);
> >  	}
> >  	return ret;
> > @@ -883,11 +883,11 @@ int transfer_args_to_stack(struct linux_binprm 
*bprm,
> >  
> >  	for (index = MAX_ARG_PAGES - 1; index >= stop; index--) {
> >  		unsigned int offset = index == stop ? bprm->p & 
~PAGE_MASK : 0;
> > -		char *src = kmap(bprm->page[index]) + offset;
> > +		char *src = kmap_local_page(bprm->page[index]) + 
offset;
> >  		sp -= PAGE_SIZE - offset;
> >  		if (copy_to_user((void *) sp, src, PAGE_SIZE - offset) 
!= 0)
> >  			ret = -EFAULT;
> > -		kunmap(bprm->page[index]);
> > +		kunmap_local(src);
> >  		if (ret)
> >  			goto out;
> >  	}
> > @@ -1680,13 +1680,13 @@ int remove_arg_zero(struct linux_binprm *bprm)
> >  			ret = -EFAULT;
> >  			goto out;
> >  		}
> > -		kaddr = kmap_atomic(page);
> > +		kaddr = kmap_local_page(page);
> >  
> >  		for (; offset < PAGE_SIZE && kaddr[offset];
> >  				offset++, bprm->p++)
> >  			;
> >  
> > -		kunmap_atomic(kaddr);
> > +		kunmap_local(kaddr);
> >  		put_arg_page(page);
> >  	} while (offset == PAGE_SIZE);
> 

Powered by blists - more mailing lists