From: Jan Beulich <jbeulich@suse.com>
To: Andrew Cooper <amc96@cam.ac.uk>
Cc: Xen-devel <xen-devel@lists.xenproject.org>,
"George Dunlap" <George.Dunlap@eu.citrix.com>,
"Ian Jackson" <iwj@xenproject.org>,
"Stefano Stabellini" <sstabellini@kernel.org>,
"Wei Liu" <wl@xen.org>, "Julien Grall" <julien@xen.org>,
"Paul Durrant" <paul@xen.org>,
"Michał Leszczyński" <michal.leszczynski@cert.pl>,
"Hubert Jasudowicz" <hubert.jasudowicz@cert.pl>,
"Tamas K Lengyel" <tamas@tklengyel.com>
Subject: Re: [PATCH v2 09/11] xen/memory: Fix mapping grant tables with XENMEM_acquire_resource
Date: Tue, 12 Jan 2021 09:39:59 +0100 [thread overview]
Message-ID: <3a661aa1-25c3-52a7-6d18-4ad5e2d9d0f7@suse.com> (raw)
In-Reply-To: <0462d837-947b-30a8-21ab-5a9a6b9ca4e1@cam.ac.uk>
On 11.01.2021 21:05, Andrew Cooper wrote:
> On 28/09/2020 10:37, Jan Beulich wrote:
>> On 22.09.2020 20:24, Andrew Cooper wrote:
>>> --- a/xen/common/compat/memory.c
>>> +++ b/xen/common/compat/memory.c
>>> @@ -636,15 +662,45 @@ int compat_memory_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) compat)
>>> compat_frame_list[i] = frame;
>>> }
>>>
>>> - if ( __copy_to_compat_offset(cmp.mar.frame_list, 0,
>>> - compat_frame_list,
>>> - cmp.mar.nr_frames) )
>>> + if ( __copy_to_compat_offset(
>>> + cmp.mar.frame_list, start_extent,
>>> + compat_frame_list, done) )
>>> return -EFAULT;
>>> }
>>> - break;
>>> +
>>> + start_extent += done;
>>> +
>>> + /* Completely done. */
>>> + if ( start_extent == cmp.mar.nr_frames )
>>> + break;
>>> +
>>> + /*
>>> + * Done a "full" batch, but we were limited by space in the xlat
>>> + * area. Go around the loop again without necesserily returning
>>> + * to guest context.
>>> + */
>>> + if ( done == nat.mar->nr_frames )
>>> + {
>>> + split = 1;
>>> + break;
>>> + }
>>> +
>>> + /* Explicit continuation request from a higher level. */
>>> + if ( done < nat.mar->nr_frames )
>>> + return hypercall_create_continuation(
>>> + __HYPERVISOR_memory_op, "ih",
>>> + op | (start_extent << MEMOP_EXTENT_SHIFT), compat);
>>> +
>>> + /*
>>> + * Well... Somethings gone wrong with the two levels of chunking.
>>> + * My condolences to whomever next has to debug this mess.
>>> + */
>> Any suggestion how to overcome this "mess"?
>
> The double level of array handling is what makes it so complicated.
> There are enough cases in compat_memory_op() alone which can't
>
> We've got two cases in practice. A singleton object needing conversion,
> or a large array of them. I'm quite certain we'd have less code and
> less complexity by having copy_$OJBECT_{to,from}_guest() helpers which
> dealt with compat internally as appropriate.
>
> We don't care about the performance of 32bit hypercalls, but not doing
> batch conversions of 1020/etc objects in the compat layer will probably
> result in better performance overall, as we don't throw away the work as
> we batch things at smaller increments higher up the stack.
I've put this on my todo list to investigate. Maybe nowadays we
really don't care all this much about 32-bit hypercall performance.
The picture was surely different when the compat layer was
introduced.
>>> --- a/xen/common/grant_table.c
>>> +++ b/xen/common/grant_table.c
>>> @@ -4105,6 +4105,9 @@ int gnttab_acquire_resource(
>>> for ( i = 0; i < nr_frames; ++i )
>>> mfn_list[i] = virt_to_mfn(vaddrs[frame + i]);
>>>
>>> + /* Success. Passed nr_frames back to the caller. */
>> Nit: "Pass"?
>
> We have already passed them back to the caller. "Pass" is the wrong
> tense to use.
Hmm, interesting view. I personally wouldn't consider the
passing back to have completed before we've returned.
Jan
next prev parent reply other threads:[~2021-01-12 8:40 UTC|newest]
Thread overview: 42+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-09-22 18:24 [PATCH v2 00/11] Multiple fixes to XENMEM_acquire_resource Andrew Cooper
2020-09-22 18:24 ` [PATCH v2 01/11] xen/memory: Introduce CONFIG_ARCH_ACQUIRE_RESOURCE Andrew Cooper
2020-09-22 18:24 ` [PATCH v2 02/11] xen/gnttab: Rework resource acquisition Andrew Cooper
2020-09-24 9:51 ` Paul Durrant
2021-01-11 21:22 ` Andrew Cooper
2021-01-12 8:23 ` Jan Beulich
2021-01-12 20:06 ` Andrew Cooper
2021-01-12 8:29 ` Paul Durrant
2020-09-25 13:17 ` Jan Beulich
2021-01-11 21:22 ` Andrew Cooper
2021-01-12 8:15 ` Jan Beulich
2021-01-12 18:11 ` Andrew Cooper
2020-09-22 18:24 ` [PATCH v2 03/11] xen/memory: Fix compat XENMEM_acquire_resource for size requests Andrew Cooper
2020-09-22 18:24 ` [PATCH v2 04/11] xen/memory: Fix acquire_resource size semantics Andrew Cooper
2020-09-24 10:06 ` Paul Durrant
2020-09-24 10:57 ` Andrew Cooper
2020-09-24 11:04 ` Paul Durrant
2020-09-25 15:56 ` Jan Beulich
2020-09-22 18:24 ` [PATCH v2 05/11] tools/foreignmem: Support querying the size of a resource Andrew Cooper
2021-01-08 17:52 ` Andrew Cooper
2021-01-11 10:50 ` Roger Pau Monné
2021-01-11 15:00 ` Andrew Cooper
2021-01-11 15:26 ` [PATCH v3 " Andrew Cooper
2021-01-11 15:54 ` Roger Pau Monné
2020-09-22 18:24 ` [PATCH v2 06/11] xen/memory: Clarify the XENMEM_acquire_resource ABI description Andrew Cooper
2020-09-24 10:08 ` Paul Durrant
2020-09-22 18:24 ` [PATCH v2 07/11] xen/memory: Improve compat XENMEM_acquire_resource handling Andrew Cooper
2020-09-24 10:16 ` Paul Durrant
2020-09-28 9:09 ` Jan Beulich
2021-01-08 18:57 ` Andrew Cooper
2021-01-11 14:25 ` Jan Beulich
2020-09-22 18:24 ` [PATCH v2 08/11] xen/memory: Indent part of acquire_resource() Andrew Cooper
2020-09-24 10:36 ` Paul Durrant
2020-09-22 18:24 ` [PATCH v2 09/11] xen/memory: Fix mapping grant tables with XENMEM_acquire_resource Andrew Cooper
2020-09-24 10:47 ` Paul Durrant
2021-01-08 19:36 ` Andrew Cooper
2020-09-28 9:37 ` Jan Beulich
2021-01-11 20:05 ` Andrew Cooper
2021-01-11 22:36 ` Andrew Cooper
2021-01-12 8:39 ` Jan Beulich [this message]
2020-09-22 18:24 ` [PATCH v2 10/11] TESTING dom0 Andrew Cooper
2020-09-22 18:24 ` [PATCH v2 11/11] TESTING XTF Andrew Cooper
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 \
--in-reply-to=3a661aa1-25c3-52a7-6d18-4ad5e2d9d0f7@suse.com \
--to=jbeulich@suse.com \
--cc=George.Dunlap@eu.citrix.com \
--cc=amc96@cam.ac.uk \
--cc=hubert.jasudowicz@cert.pl \
--cc=iwj@xenproject.org \
--cc=julien@xen.org \
--cc=michal.leszczynski@cert.pl \
--cc=paul@xen.org \
--cc=sstabellini@kernel.org \
--cc=tamas@tklengyel.com \
--cc=wl@xen.org \
--cc=xen-devel@lists.xenproject.org \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.