From: Julien Grall <julien@xen.org>
To: Bertrand Marquis <bertrand.marquis@arm.com>,
xen-devel@lists.xenproject.org
Cc: nd@arm.com, Volodymyr Babchuk <Volodymyr_Babchuk@epam.com>,
Stefano Stabellini <sstabellini@kernel.org>
Subject: Re: [PATCH 2/2] xen/arm: Support runstate crossing pages
Date: Fri, 12 Jun 2020 13:14:56 +0100 [thread overview]
Message-ID: <0ea51410-050b-58a6-806a-b175f534852f@xen.org> (raw)
In-Reply-To: <b4843bd234d4ece4f843bc636071106746abb3b5.1591806713.git.bertrand.marquis@arm.com>
Hi,
On 11/06/2020 12:58, Bertrand Marquis wrote:
> Add support for runstate area register with the structure crossing pages
Well, this has always been supported until your previous patch. In
general, we try to not break thing in a middle of the series so we can
still bisect it.
I think this patch can be simplified a lot by mapping the two page
contiguously (see my previous answer). With that it would be feasible to
fold this patch in #1.
> The code is storing up to 2 pages reference during the hypercall.
> During a context switch, the code is computing where the
> state_entry_time is and is breaking the memcpy in 2 parts when it is
> required.
>
> Signed-off-by: Bertrand Marquis <bertrand.marquis@arm.com>
> ---
> xen/arch/arm/domain.c | 101 +++++++++++++++++++++++++----------
> xen/include/asm-arm/domain.h | 5 +-
> 2 files changed, 76 insertions(+), 30 deletions(-)
>
> diff --git a/xen/arch/arm/domain.c b/xen/arch/arm/domain.c
> index 739059234f..d847cb00f2 100644
> --- a/xen/arch/arm/domain.c
> +++ b/xen/arch/arm/domain.c
> @@ -280,11 +280,16 @@ void arch_cleanup_runstate_guest(struct vcpu *v)
> {
> spin_lock(&v->arch.runstate_guest.lock);
>
> - /* cleanup previous page if any */
> - if ( v->arch.runstate_guest.page )
> + /* cleanup previous pages if any */
> + if ( v->arch.runstate_guest.page[0] )
> {
> - put_page_and_type(v->arch.runstate_guest.page);
> - v->arch.runstate_guest.page = NULL;
> + put_page_and_type(v->arch.runstate_guest.page[0]);
> + v->arch.runstate_guest.page[0] = NULL;
> + if ( v->arch.runstate_guest.page[1] )
> + {
> + put_page_and_type(v->arch.runstate_guest.page[1]);
> + v->arch.runstate_guest.page[1] = NULL;
> + }
I think this can be moved outside of the first if.
> v->arch.runstate_guest.offset = 0;
> }
>
> @@ -294,26 +299,25 @@ void arch_cleanup_runstate_guest(struct vcpu *v)
> int arch_setup_runstate_guest(struct vcpu *v,
> struct vcpu_register_runstate_memory_area area)
> {
> - struct page_info *page;
> + struct page_info *page[2] = {NULL,NULL};
I don't think you need the temporary variable. You can directly update
page[0] as it is protected by the lock. The nice benefits is you could
take advantage of a common helper to cleanup and reduce the complexity
of the code.
> unsigned offset;
>
> spin_lock(&v->arch.runstate_guest.lock);
>
> - /* cleanup previous page if any */
> - if ( v->arch.runstate_guest.page )
> + /* cleanup previous pages if any */
> + if ( v->arch.runstate_guest.page[0] )
> {
> - put_page_and_type(v->arch.runstate_guest.page);
> - v->arch.runstate_guest.page = NULL;
> + put_page_and_type(v->arch.runstate_guest.page[0]);
> + v->arch.runstate_guest.page[0] = NULL;
> + if ( v->arch.runstate_guest.page[1] )
> + {
> + put_page_and_type(v->arch.runstate_guest.page[1]);
> + v->arch.runstate_guest.page[1] = NULL;
> + }
> v->arch.runstate_guest.offset = 0;
> }
>
> offset = ((vaddr_t)area.addr.v) & ~PAGE_MASK;
> - if ( offset > (PAGE_SIZE - sizeof(struct vcpu_runstate_info)) )
> - {
> - spin_unlock(&v->arch.runstate_guest.lock);
> - gprintk(XENLOG_DEBUG, "Runstate is crossing pages\n");
> - return -EINVAL;
> - }
>
> /* provided address must be aligned to a 64bit */
> if ( offset % alignof(struct vcpu_runstate_info) )
> @@ -323,15 +327,30 @@ int arch_setup_runstate_guest(struct vcpu *v,
> return -EINVAL;
> }
>
> - page = get_page_from_gva(v, (vaddr_t)area.addr.v, GV2M_WRITE);
> - if ( !page )
> + page[0] = get_page_from_gva(v, (vaddr_t)area.addr.v, GV2M_WRITE);
> + if ( !page[0] )
> {
> spin_unlock(&v->arch.runstate_guest.lock);
> gprintk(XENLOG_DEBUG, "Runstate pointer is not mapped\n");
> return -EINVAL;
> }
>
> - v->arch.runstate_guest.page = page;
> + if ( offset > (PAGE_SIZE - sizeof(struct vcpu_runstate_info)) )
> + {
> + /* guest area is crossing pages */
> + page[1] = get_page_from_gva(v, (vaddr_t)area.addr.v + PAGE_SIZE,
> + GV2M_WRITE);
> + if ( !page[1] )
> + {
> + put_page_and_type(v->arch.runstate_guest.page[0]);
v->arch.runstate_guest.page[0] would be NULL as you set it afterwards.
So you want to set v->arch.runstate_guest.page[0] beforehand.
Cheers,
--
Julien Grall
next prev parent reply other threads:[~2020-06-12 12:15 UTC|newest]
Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-06-11 11:58 [PATCH 0/2] xen/arm: Convert runstate address during hypcall Bertrand Marquis
2020-06-11 11:58 ` [PATCH 1/2] " Bertrand Marquis
2020-06-11 18:16 ` Stefano Stabellini
2020-06-11 18:24 ` Julien Grall
2020-06-11 18:50 ` Stefano Stabellini
2020-06-11 19:38 ` Julien Grall
2020-06-12 1:09 ` Stefano Stabellini
2020-06-12 8:13 ` Bertrand Marquis
2020-06-13 0:24 ` Stefano Stabellini
2020-06-15 14:09 ` Bertrand Marquis
2020-06-15 20:30 ` Stefano Stabellini
2020-06-15 20:44 ` Julien Grall
2020-06-12 9:53 ` Julien Grall
2020-06-13 0:24 ` Stefano Stabellini
2020-06-12 8:07 ` Bertrand Marquis
2020-06-12 10:53 ` Julien Grall
2020-06-12 14:13 ` Bertrand Marquis
2020-06-12 19:56 ` Julien Grall
2020-06-12 16:51 ` Bertrand Marquis
2020-06-12 20:31 ` Julien Grall
2020-06-15 14:01 ` Bertrand Marquis
2020-06-11 11:58 ` [PATCH 2/2] xen/arm: Support runstate crossing pages Bertrand Marquis
2020-06-12 1:10 ` Stefano Stabellini
2020-06-12 11:37 ` Julien Grall
2020-06-12 12:14 ` Julien Grall [this message]
2020-06-12 16:13 ` Bertrand Marquis
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=0ea51410-050b-58a6-806a-b175f534852f@xen.org \
--to=julien@xen.org \
--cc=Volodymyr_Babchuk@epam.com \
--cc=bertrand.marquis@arm.com \
--cc=nd@arm.com \
--cc=sstabellini@kernel.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).