public inbox for [email protected]
 help / color / mirror / Atom feed
From: Helge Deller <[email protected]>
To: Jens Axboe <[email protected]>,
	[email protected], [email protected],
	John David Anglin <[email protected]>
Subject: Re: [PATCH v2] io_uring: Adjust mapping wrt architecture aliasing requirements
Date: Thu, 16 Feb 2023 17:33:15 +0100	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>

On 2/16/23 17:11, Jens Axboe wrote:
> On 2/16/23 1:09?AM, Helge Deller wrote:
>> Some architectures have memory cache aliasing requirements (e.g. parisc)
>> if memory is shared between userspace and kernel. This patch fixes the
>> kernel to return an aliased address when asked by userspace via mmap().
>>
>> Signed-off-by: Helge Deller <[email protected]>
>> ---
>> v2: Do not allow to map to a user-provided addresss. This forces
>> programs to write portable code, as usually on x86 mapping to any
>> address will succeed, while it will fail for most provided address if
>> used on stricter architectures.
>>
>> diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
>> index 862e05e6691d..01fe7437a071 100644
>> --- a/io_uring/io_uring.c
>> +++ b/io_uring/io_uring.c
>> @@ -72,6 +72,7 @@
>>   #include <linux/io_uring.h>
>>   #include <linux/audit.h>
>>   #include <linux/security.h>
>> +#include <asm/shmparam.h>
>>
>>   #define CREATE_TRACE_POINTS
>>   #include <trace/events/io_uring.h>
>> @@ -3059,6 +3060,54 @@ static __cold int io_uring_mmap(struct file *file, struct vm_area_struct *vma)
>>   	return remap_pfn_range(vma, vma->vm_start, pfn, sz, vma->vm_page_prot);
>>   }
>>
>> +static unsigned long io_uring_mmu_get_unmapped_area(struct file *filp,
>> +			unsigned long addr, unsigned long len,
>> +			unsigned long pgoff, unsigned long flags)
>> +{
>> +	const unsigned long mmap_end = arch_get_mmap_end(addr, len, flags);
>> +	struct vm_unmapped_area_info info;
>> +	void *ptr;
>> +
>> +	/*
>> +	 * Do not allow to map to user-provided address to avoid breaking the
>> +	 * aliasing rules. Userspace is not able to guess the offset address of
>> +	 * kernel kmalloc()ed memory area.
>> +	 */
>> +	if (addr)
>> +		return -EINVAL;
>
> Can we relax this so that if the address is correctly aligned, it will
> allow it?

My previous patch had it relaxed, but after some more thoughts I removed
it in this v2-version again.

The idea behind it is good, but I see a huge disadvantage in allowing
correctly aligned addresses: People develop their code usually on x86
which has no such alignment requirements, as it just needs to be PAGE_SIZE aligned.
So their code will always work fine on x86, but as soon as the same code
is built on other platforms it will break. As you know, on parisc it's pure luck
if the program chooses an address which is correctly aligned.
I'm one of the debian maintainers for parisc, and I've seen similiar
mmap-issues in other programs as well. Everytime I've found it to be wrong,
you have to explain to the developers what's wrong and sometimes it's
not easy to fix it.
So, if we can educate people from assuming their code to be correct, I think
we can save a lot of additional work afterwards.
That said, I think it's better to be strict now, unless someone comes
up with a really good reason why it needs to be less strict.

> The reported issue with sqpoll-cancel-hang.t is due to it
> crashing because it's a weird syzbot thing that does mmap() with
> MAP_FIXED and an address given.

Ok, but nevertheless I think it's better to be strict.

Helge

  reply	other threads:[~2023-02-16 16:33 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-02-16  8:09 [PATCH v2] io_uring: Adjust mapping wrt architecture aliasing requirements Helge Deller
2023-02-16 16:11 ` Jens Axboe
2023-02-16 16:33   ` Helge Deller [this message]
2023-02-16 16:46     ` Jens Axboe
2023-02-16 17:52       ` Helge Deller
2023-02-16 18:00         ` Jens Axboe
2023-06-27 14:14 ` Jiri Slaby
2023-06-27 19:24   ` Helge Deller

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox