From: Oleksandr <olekstysh@gmail.com>
To: Julien Grall <julien@xen.org>
Cc: xen-devel@lists.xenproject.org,
Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>,
Stefano Stabellini <sstabellini@kernel.org>,
Volodymyr Babchuk <Volodymyr_Babchuk@epam.com>,
Julien Grall <julien.grall@arm.com>
Subject: Re: [PATCH V5 22/22] xen/arm: Add mapcache invalidation handling
Date: Thu, 28 Jan 2021 15:12:17 +0200 [thread overview]
Message-ID: <9d784a7e-2669-965c-a01d-cca3452b8164@gmail.com> (raw)
In-Reply-To: <1818a998-fe85-b24d-2141-a755a3e74623@xen.org>
On 28.01.21 11:55, Julien Grall wrote:
> Hi Oleksandr,
Hi Julien
>
> On 25/01/2021 19:08, Oleksandr Tyshchenko wrote:
>> From: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>
>>
>> We need to send mapcache invalidation request to qemu/demu everytime
>> the page gets removed from a guest.
>>
>> At the moment, the Arm code doesn't explicitely remove the existing
>> mapping before inserting the new mapping. Instead, this is done
>> implicitely by __p2m_set_entry().
>>
>> So we need to recognize a case when old entry is a RAM page *and*
>> the new MFN is different in order to set the corresponding flag.
>
> AFAICT, this sentence doesn't match the code. Instead, you will
> request the mapcache invalidation whenever the "freed" entry contains
> some RAM page.
>
> I am fine with the approach used in the code. However, it would be
> good that the commit message reflects the code.
ok, will update it.
>
>
>> The most suitable place to do this is p2m_free_entry(), there
>> we can find the correct leaf type. The invalidation request
>> will be sent in do_trap_hypercall() later on.
>>
>> Taking into the account the following the do_trap_hypercall()
>> is the best place to send invalidation request:
>> - The only way a guest can modify its P2M on Arm is via an hypercall
>> - When sending the invalidation request, the vCPU will be blocked
>> until all the IOREQ servers have acknowledged the invalidation
>>
>> Signed-off-by: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>
>> CC: Julien Grall <julien.grall@arm.com>
>> Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
>> [On Arm only]
>> Tested-by: Wei Chen <Wei.Chen@arm.com>
>>
>> ---
>> Please note, this is a split/cleanup/hardening of Julien's PoC:
>> "Add support for Guest IO forwarding to a device emulator"
>>
>> ***
>> Please note, this patch depends on the following which is
>> on review:
>> https://patchwork.kernel.org/patch/11803383/
>
> I think we can defer this post this series.
ok
>
>
>>
>> This patch is on par with x86 code (whether it is buggy or not).
>> If there is a need to improve/harden something, this can be done on
>> a follow-up.
>> ***
>>
>> Changes V1 -> V2:
>> - new patch, some changes were derived from (+ new explanation):
>> xen/ioreq: Make x86's invalidate qemu mapcache handling common
>> - put setting of the flag into __p2m_set_entry()
>> - clarify the conditions when the flag should be set
>> - use domain_has_ioreq_server()
>> - update do_trap_hypercall() by adding local variable
>>
>> Changes V2 -> V3:
>> - update patch description
>> - move check to p2m_free_entry()
>> - add a comment
>> - use "curr" instead of "v" in do_trap_hypercall()
>>
>> Changes V3 -> V4:
>> - update patch description
>> - re-order check in p2m_free_entry() to call
>> domain_has_ioreq_server()
>> only if p2m->domain == current->domain
>> - add a comment in do_trap_hypercall()
>>
>> Changes V4 -> V5:
>> - add Stefano's R-b
>> - update comment in do_trap_hypercall()
>> ---
>> ---
>> xen/arch/arm/p2m.c | 25 +++++++++++++++++--------
>> xen/arch/arm/traps.c | 20 +++++++++++++++++---
>> 2 files changed, 34 insertions(+), 11 deletions(-)
>>
>> diff --git a/xen/arch/arm/p2m.c b/xen/arch/arm/p2m.c
>> index d41c4fa..26acb95d 100644
>> --- a/xen/arch/arm/p2m.c
>> +++ b/xen/arch/arm/p2m.c
>> @@ -1,6 +1,7 @@
>> #include <xen/cpu.h>
>> #include <xen/domain_page.h>
>> #include <xen/iocap.h>
>> +#include <xen/ioreq.h>
>> #include <xen/lib.h>
>> #include <xen/sched.h>
>> #include <xen/softirq.h>
>> @@ -749,17 +750,25 @@ static void p2m_free_entry(struct p2m_domain *p2m,
>> if ( !p2m_is_valid(entry) )
>> return;
>> - /* Nothing to do but updating the stats if the entry is a
>> super-page. */
>> - if ( p2m_is_superpage(entry, level) )
>> + if ( p2m_is_superpage(entry, level) || (level == 3) )
>> {
>> - p2m->stats.mappings[level]--;
>> - return;
>> - }
>> +#ifdef CONFIG_IOREQ_SERVER
>> + /*
>> + * If this gets called (non-recursively) then either the entry
>
> I am not sure why you specify "non-recursively". You still want to
> invalidate the mapcache if we replaced a table with a superpage mapping.
yes agree, sorry I don't quite remember why I decided to put emphasis on
this word. I will just remove it.
>
>
>> + * was replaced by an entry with a different base (valid
>> case) or
>> + * the shattering of a superpage was failed (error case).
>
> s/was/has/ I think.
ok, I assume this applies for the second part of the sentence.
>
>> + * So, at worst, the spurious mapcache invalidation might be
>> sent.
>> + */
>> + if ( (p2m->domain == current->domain) &&
>> + domain_has_ioreq_server(p2m->domain) &&
>> + p2m_is_ram(entry.p2m.type) )
>> + p2m->domain->mapcache_invalidate = true;
>> +#endif
>> - if ( level == 3 )
>> - {
>> p2m->stats.mappings[level]--;
>> - p2m_put_l3_page(entry);
>> + /* Nothing to do if the entry is a super-page. */
>> + if ( level == 3 )
>> + p2m_put_l3_page(entry);
>> return;
>> }
>> diff --git a/xen/arch/arm/traps.c b/xen/arch/arm/traps.c
>> index 4cdd343..64b740b 100644
>> --- a/xen/arch/arm/traps.c
>> +++ b/xen/arch/arm/traps.c
>> @@ -1443,6 +1443,7 @@ static void do_trap_hypercall(struct
>> cpu_user_regs *regs, register_t *nr,
>> const union hsr hsr)
>> {
>> arm_hypercall_fn_t call = NULL;
>> + struct vcpu *curr = current;
>> BUILD_BUG_ON(NR_hypercalls < ARRAY_SIZE(arm_hypercall_table) );
>> @@ -1459,7 +1460,7 @@ static void do_trap_hypercall(struct
>> cpu_user_regs *regs, register_t *nr,
>> return;
>> }
>> - current->hcall_preempted = false;
>> + curr->hcall_preempted = false;
>> perfc_incra(hypercalls, *nr);
>> call = arm_hypercall_table[*nr].fn;
>> @@ -1472,7 +1473,7 @@ static void do_trap_hypercall(struct
>> cpu_user_regs *regs, register_t *nr,
>> HYPERCALL_RESULT_REG(regs) = call(HYPERCALL_ARGS(regs));
>> #ifndef NDEBUG
>> - if ( !current->hcall_preempted )
>> + if ( !curr->hcall_preempted )
>> {
>> /* Deliberately corrupt parameter regs used by this
>> hypercall. */
>> switch ( arm_hypercall_table[*nr].nr_args ) {
>> @@ -1489,8 +1490,21 @@ static void do_trap_hypercall(struct
>> cpu_user_regs *regs, register_t *nr,
>> #endif
>> /* Ensure the hypercall trap instruction is re-executed. */
>> - if ( current->hcall_preempted )
>> + if ( curr->hcall_preempted )
>> regs->pc -= 4; /* re-execute 'hvc #XEN_HYPERCALL_TAG' */
>> +
>> +#ifdef CONFIG_IOREQ_SERVER
>> + /*
>> + * We call ioreq_signal_mapcache_invalidate from
>> do_trap_hypercall()
>> + * because the only way a guest can modify its P2M on Arm is via an
>> + * hypercall.
>> + * Note that sending the invalidation request causes the vCPU to
>> block
>> + * until all the IOREQ servers have acknowledged the invalidation.
>> + */
>> + if ( unlikely(curr->domain->mapcache_invalidate) &&
>> + test_and_clear_bool(curr->domain->mapcache_invalidate) )
>> + ioreq_signal_mapcache_invalidate();
>> +#endif
>> }
>> void arch_hypercall_tasklet_result(struct vcpu *v, long res)
>>
>
> Cheers,
>
--
Regards,
Oleksandr Tyshchenko
next prev parent reply other threads:[~2021-01-28 13:12 UTC|newest]
Thread overview: 93+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-25 19:08 [PATCH V5 00/22] IOREQ feature (+ virtio-mmio) on Arm Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 01/22] x86/ioreq: Prepare IOREQ feature for making it common Oleksandr Tyshchenko
2021-01-27 16:48 ` Jan Beulich
2021-01-25 19:08 ` [PATCH V5 02/22] x86/ioreq: Add IOREQ_STATUS_* #define-s and update code for moving Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 03/22] x86/ioreq: Provide out-of-line wrapper for the handle_mmio() Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 04/22] xen/ioreq: Make x86's IOREQ feature common Oleksandr Tyshchenko
2021-01-25 23:13 ` Julien Grall
2021-01-25 23:40 ` Oleksandr
2021-01-27 18:43 ` Julien Grall
2021-01-27 20:22 ` Oleksandr
2021-01-27 20:46 ` Stefano Stabellini
2021-01-28 11:01 ` Oleksandr
2021-01-28 11:21 ` Jan Beulich
2021-01-28 12:49 ` Oleksandr
2021-01-28 17:13 ` Stefano Stabellini
2021-01-27 16:58 ` Jan Beulich
2021-01-27 20:14 ` Oleksandr
2021-01-28 8:06 ` Jan Beulich
2021-01-28 11:16 ` Oleksandr
2021-01-28 11:24 ` Jan Beulich
2021-01-25 19:08 ` [PATCH V5 05/22] xen/ioreq: Make x86's hvm_ioreq_needs_completion() common Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 06/22] xen/ioreq: Make x86's hvm_mmio_first(last)_byte() common Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 07/22] xen/ioreq: Make x86's hvm_ioreq_(page/vcpu/server) structs common Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 08/22] xen/ioreq: Move x86's ioreq_server to struct domain Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 09/22] xen/ioreq: Make x86's IOREQ related dm-op handling common Oleksandr Tyshchenko
2021-01-26 13:31 ` Paul Durrant
2021-01-28 10:52 ` Jan Beulich
2021-01-28 12:06 ` Oleksandr
2021-01-28 13:18 ` Jan Beulich
2021-01-28 20:01 ` Oleksandr
2021-01-25 19:08 ` [PATCH V5 10/22] xen/ioreq: Move x86's io_completion/io_req fields to struct vcpu Oleksandr Tyshchenko
2021-01-28 13:41 ` Julien Grall
2021-01-28 13:53 ` Jan Beulich
2021-01-28 14:21 ` Julien Grall
2021-01-28 14:36 ` Jan Beulich
2021-01-28 14:49 ` Andrew Cooper
2021-01-28 14:51 ` Ian Jackson
2021-01-28 14:54 ` Jan Beulich
2021-01-28 16:15 ` Andrew Cooper
2021-01-28 15:01 ` Julien Grall
2021-01-28 14:05 ` Oleksandr
2021-01-25 19:08 ` [PATCH V5 11/22] xen/mm: Make x86's XENMEM_resource_ioreq_server handling common Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 12/22] xen/ioreq: Remove "hvm" prefixes from involved function names Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 13/22] xen/ioreq: Use guest_cmpxchg64() instead of cmpxchg() Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 14/22] arm/ioreq: Introduce arch specific bits for IOREQ/DM features Oleksandr Tyshchenko
2021-01-25 20:19 ` Stefano Stabellini
2021-01-26 9:15 ` Jan Beulich
2021-01-27 9:54 ` Oleksandr
2021-01-27 10:15 ` Jan Beulich
2021-01-27 17:01 ` Julien Grall
2021-01-27 17:04 ` Jan Beulich
2021-01-27 18:33 ` Julien Grall
2021-01-27 19:20 ` Oleksandr
2021-01-28 9:40 ` Julien Grall
2021-01-28 11:33 ` Oleksandr
2021-01-28 13:39 ` Oleksandr
2021-01-28 14:29 ` Oleksandr
2021-01-28 14:41 ` Julien Grall
2021-01-28 14:52 ` Oleksandr
2021-01-28 15:08 ` Julien Grall
2021-01-28 17:50 ` Oleksandr
2021-01-25 19:08 ` [PATCH V5 15/22] xen/arm: Call vcpu_ioreq_handle_completion() in check_for_vcpu_work() Oleksandr Tyshchenko
2021-01-27 14:49 ` Julien Grall
2021-01-27 15:56 ` Oleksandr
2021-01-27 20:34 ` Stefano Stabellini
2021-01-25 19:08 ` [PATCH V5 16/22] xen/mm: Handle properly reference in set_foreign_p2m_entry() on Arm Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 17/22] xen/ioreq: Introduce domain_has_ioreq_server() Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 18/22] xen/dm: Introduce xendevicemodel_set_irq_level DM op Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 19/22] xen/arm: io: Abstract sign-extension Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 20/22] xen/arm: io: Harden sign extension check Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 21/22] xen/ioreq: Make x86's send_invalidate_req() common Oleksandr Tyshchenko
2021-01-25 19:08 ` [PATCH V5 22/22] xen/arm: Add mapcache invalidation handling Oleksandr Tyshchenko
2021-01-28 9:55 ` Julien Grall
2021-01-28 13:12 ` Oleksandr [this message]
2021-01-27 16:43 ` [PATCH V5 00/22] IOREQ feature (+ virtio-mmio) on Arm Julien Grall
2021-01-27 16:50 ` Oleksandr
2021-01-27 17:33 ` Julien Grall
2021-01-27 17:37 ` Oleksandr
2021-01-27 17:42 ` Julien Grall
2021-01-27 17:45 ` Oleksandr
2021-01-28 14:37 ` Oleksandr
2021-01-28 15:14 ` Julien Grall
2021-01-28 17:55 ` Oleksandr
2021-01-28 16:11 ` Julien Grall
2021-01-28 17:24 ` Stefano Stabellini
2021-01-28 17:44 ` Julien Grall
2021-01-28 18:10 ` Andrew Cooper
2021-01-28 18:16 ` Julien Grall
2021-01-28 18:21 ` Julien Grall
2021-01-28 20:10 ` Andrew Cooper
2021-01-28 21:19 ` Julien Grall
2021-01-28 19:44 ` Oleksandr
2021-01-29 1:15 ` Oleksandr
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=9d784a7e-2669-965c-a01d-cca3452b8164@gmail.com \
--to=olekstysh@gmail.com \
--cc=Volodymyr_Babchuk@epam.com \
--cc=julien.grall@arm.com \
--cc=julien@xen.org \
--cc=oleksandr_tyshchenko@epam.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).