Skip to contentSkip to conversation

splice: Fix corruption of spliced data after splice() returns

RFC PATCH In review Alejandro Colomar 1 participant 2 messages Jul 20, 2023 – Jul 28, 2023 1 message not in the archive Man-pages: manual pages for Linux -- Sections 2, 3, 4, 5, and 7

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

Alejandro Colomar

author maintainer
#1 lore raw

RFC PATCH 1/4 splice: Fix corruption of spliced data after splice() returns

Hi Randy!

Thanks! I'll FWD to the linux-man@ mailing list too.

Cheers,
Alex

-------- Forwarded Message --------
Subject: Fwd: [RFC PATCH 1/4] splice: Fix corruption of spliced data after splice() returns
Date: Wed, 19 Jul 2023 17:36:03 -0700
From: Randy Dunlap <[email protected]>
To: Alejandro Colomar <[email protected]>, Michael Kerrisk <[email protected]>

FYI:

-------- Forwarded Message --------
Subject: Re: [RFC PATCH 1/4] splice: Fix corruption of spliced data after splice() returns
Date: Wed, 19 Jul 2023 17:00:17 -0700
From: Linus Torvalds <[email protected]>
To: Matt Whitlock <[email protected]>
CC: Matthew Wilcox <[email protected]>, Miklos Szeredi <[email protected]>, David Howells <[email protected]>, [email protected], Dave Chinner <[email protected]>, Jens Axboe <[email protected]>, [email protected], [email protected], [email protected], Christoph Hellwig <[email protected]>, [email protected]

Quoted 5 lines · Matt Whitlock

Then that is my request. This entire complaint/discussion/argument would have been avoided if splice(2) had contained a sentence like this one from sendfile(2):

Quoted 7 lines

"If out_fd refers to a socket or pipe with zero-copy support, callers must ensure the transferred portions of the file referred to by in_fd remain unmodified until the reader on the other end of out_fd has consumed the transferred data."

That is a clear warning of the perils of the implementation under the hood, and it could/should be copied, more or less verbatim, to splice(2).

Ack. Internally in the kernel, the two really have always been more or less of intermingled.

In fact, I think splice()/sendfile()/tee() could - and maybe should - actually be a single man-page to make it clear that they are all facets of the same thing.

The issues with TCP_CORK exist for splice too, for example, for exactly the same reasons.

And while SPLICE_F_MORE exists, it only deals with multiple splice() calls, it doesn't deal with the "I wrote a header before I even started using splice()" case that is the one that is mentioned for sendfile().

Or course, technically TCP_CORK exists for plain write() use as well, but there the portable and historical fix is simply to use writev() and send it all in one go.

So it's hopefully only when you use sendfile() and splice() that you end up with "oh, but I have multiple different *kinds* of sources, and I want to cork things until I've dealt with them all".

Linus

author maintainer
#2 lore raw

Hi Randy, Linus,

Quoted 11 lines · Linus Torvalds

Oh, and none of this discussion has had any man-page people associated with it.

So in order to maybe get the documentation clarified, let's add them.

Alejandro, Michael, see

https://lore.kernel.org/lkml/[email protected]/

for the original email (and thread).

Linus

Thank you both for pointing to that! I've been checking, and the splice.2 source code already has some hidden comments right after a paragraph.

(see <https://git.kernel.org/pub/scm/docs/man-pages/man-pages.git/tree/man2/splice.2#n215>)

I think we can append a paragraph like the suggested one after the one that starts in line 216. Can you clarify if the requirement of taking care of not modifying the buffer applied to just splice(2) or also tee(2) and/or vmsplice(2). BTW, if you want to send a patch, that's very welcome.

Cheers,
Alex

Quoted 62 lines · Alejandro Colomar

Hi Randy!

Thanks! I'll FWD to the linux-man@ mailing list too.

Cheers,
Alex

-------- Forwarded Message --------
Subject: Fwd: [RFC PATCH 1/4] splice: Fix corruption of spliced data after splice() returns
Date: Wed, 19 Jul 2023 17:36:03 -0700
From: Randy Dunlap <[email protected]>
To: Alejandro Colomar <[email protected]>, Michael Kerrisk <[email protected]>

FYI:

-------- Forwarded Message --------
Subject: Re: [RFC PATCH 1/4] splice: Fix corruption of spliced data after splice() returns
Date: Wed, 19 Jul 2023 17:00:17 -0700
From: Linus Torvalds <[email protected]>
To: Matt Whitlock <[email protected]>
CC: Matthew Wilcox <[email protected]>, Miklos Szeredi <[email protected]>, David Howells <[email protected]>, [email protected], Dave Chinner <[email protected]>, Jens Axboe <[email protected]>, [email protected], [email protected], [email protected], Christoph Hellwig <[email protected]>, [email protected]

Quoted 5 lines · Matt Whitlock

Then that is my request. This entire complaint/discussion/argument would have been avoided if splice(2) had contained a sentence like this one from sendfile(2):

Quoted 7 lines

"If out_fd refers to a socket or pipe with zero-copy support, callers must ensure the transferred portions of the file referred to by in_fd remain unmodified until the reader on the other end of out_fd has consumed the transferred data."

That is a clear warning of the perils of the implementation under the hood, and it could/should be copied, more or less verbatim, to splice(2).

Ack. Internally in the kernel, the two really have always been more or less of intermingled.

In fact, I think splice()/sendfile()/tee() could - and maybe should - actually be a single man-page to make it clear that they are all facets of the same thing.

The issues with TCP_CORK exist for splice too, for example, for exactly the same reasons.

And while SPLICE_F_MORE exists, it only deals with multiple splice() calls, it doesn't deal with the "I wrote a header before I even started using splice()" case that is the one that is mentioned for sendfile().

Or course, technically TCP_CORK exists for plain write() use as well, but there the portable and historical fix is simply to use writev() and send it all in one go.

So it's hopefully only when you use sendfile() and splice() that you end up with "oh, but I have multiple different *kinds* of sources, and I want to cork things until I've dealt with them all".

Linus

Signature · 2 lines
<http://www.alejandro-colomar.es/>
GPG key fingerprint: A9348594CE31283A826FBDD8D57633D441E25BB5