linux-kernel - Re: [PATCH v4 0/3] io_uring: add splice(2) support

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <189e8cb4-df3a-f12d-9b21-7134caa918bb@gmail.com>
Date:   Tue, 25 Feb 2020 01:53:55 +0300
From:   Pavel Begunkov <asml.silence@...il.com>
To:     Jens Axboe <axboe@...nel.dk>,
        Alexander Viro <viro@...iv.linux.org.uk>,
        Stefan Metzmacher <metze@...ba.org>, io-uring@...r.kernel.org,
        linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v4 0/3] io_uring: add splice(2) support

On 25/02/2020 01:51, Pavel Begunkov wrote:
> On 25/02/2020 01:34, Jens Axboe wrote:
>> On 2/24/20 8:35 AM, Jens Axboe wrote:
>>> On 2/24/20 1:32 AM, Pavel Begunkov wrote:
>>>> *on top of for-5.6 + async patches*
>>>>
>>>> Not the fastets implementation, but I'd need to stir up/duplicate
>>>> splice.c bits to do it more efficiently.
>>>>
>>>> note: rebase on top of the recent inflight patchset.
>>>
>>> Let's get this queued up, looks good to go to me. Do you have a few
>>> liburing test cases we can add for this?
>>
>> Seems to me like we have an address space issue for the off_in and
> 
> Is that a problem? From the old fixing thread loop_rw_iter() it appeared
> to me, that it's ok to pass a kernel address as a user one.
> f_op->write of some implemented through the same copy_to_user().

Either I finally need to check myself how the protection is implemented...

> 
>> off_out parameters. Why aren't we passing in pointers to these
>> and making them work like regular splice?
> 
> That's one extra copy_to_user() + copy_from_user(), which I hope to remove
> in the future. And I'm not really a fan of such API, and would prefer to give
> away such tracking to the userspace.
> 
>>
>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index 792ef01a521c..b0cfd68be8c9 100644
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -448,8 +448,8 @@ struct io_epoll {
>>  struct io_splice {
>>  	struct file			*file_out;
>>  	struct file			*file_in;
>> -	loff_t				off_out;
>> -	loff_t				off_in;
>> +	loff_t __user			*off_out;
>> +	loff_t __user			*off_in;
>>  	u64				len;
>>  	unsigned int			flags;
>>  };
>> @@ -2578,8 +2578,8 @@ static int io_splice_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
>>  		return 0;
>>  
>>  	sp->file_in = NULL;
>> -	sp->off_in = READ_ONCE(sqe->splice_off_in);
>> -	sp->off_out = READ_ONCE(sqe->off);
>> +	sp->off_in = u64_to_user_ptr(READ_ONCE(sqe->splice_off_in));
>> +	sp->off_out = u64_to_user_ptr(READ_ONCE(sqe->off));
>>  	sp->len = READ_ONCE(sqe->len);
>>  	sp->flags = READ_ONCE(sqe->splice_flags);
>>  
>> @@ -2614,7 +2614,6 @@ static int io_splice(struct io_kiocb *req, struct io_kiocb **nxt,
>>  	struct file *in = sp->file_in;
>>  	struct file *out = sp->file_out;
>>  	unsigned int flags = sp->flags & ~SPLICE_F_FD_IN_FIXED;
>> -	loff_t *poff_in, *poff_out;
>>  	long ret;
>>  
>>  	if (force_nonblock) {
>> @@ -2623,9 +2622,7 @@ static int io_splice(struct io_kiocb *req, struct io_kiocb **nxt,
>>  		flags |= SPLICE_F_NONBLOCK;
>>  	}
>>  
>> -	poff_in = (sp->off_in == -1) ? NULL : &sp->off_in;
>> -	poff_out = (sp->off_out == -1) ? NULL : &sp->off_out;
>> -	ret = do_splice(in, poff_in, out, poff_out, sp->len, flags);
>> +	ret = do_splice(in, sp->off_in, out, sp->off_out, sp->len, flags);
>>  	if (force_nonblock && ret == -EAGAIN)
>>  		return -EAGAIN;
>>  
>>
> 

-- 
Pavel Begunkov



Download attachment "signature.asc" of type "application/pgp-signature" (834 bytes)