From: Jason Wang <jasowang@redhat.com>
To: Elena Afanasova <eafanasova@gmail.com>, kvm@vger.kernel.org
Cc: stefanha@redhat.com, jag.raman@oracle.com,
elena.ufimtseva@oracle.com, pbonzini@redhat.com, mst@redhat.com,
cohuck@redhat.com, john.levon@nutanix.com
Subject: Re: [RFC v3 2/5] KVM: x86: add support for ioregionfd signal handling
Date: Tue, 9 Mar 2021 13:51:40 +0800 [thread overview]
Message-ID: <e276b54a-b2c0-c12e-fdae-22f54824ee6f@redhat.com> (raw)
In-Reply-To: <575df1656277c55f26e660b7274a7c570b448636.1613828727.git.eafanasova@gmail.com>
On 2021/2/21 8:04 下午, Elena Afanasova wrote:
> The vCPU thread may receive a signal during ioregionfd communication,
> ioctl(KVM_RUN) needs to return to userspace and then ioctl(KVM_RUN)
> must resume ioregionfd.
After a glance at the patch, I wonder can we split the patch into two?
1) sleepable iodevice which is not supported currently, probably with a
new cap?
2) ioregionfd specific codes (I wonder if it has any)
Then the sleepable iodevice could be reused by future features.
>
> Signed-off-by: Elena Afanasova <eafanasova@gmail.com>
> ---
> v3:
> - add FAST_MMIO bus support
> - move ioregion_interrupted flag to ioregion_ctx
> - reorder ioregion_ctx fields
> - rework complete_ioregion operations
> - add signal handling support for crossing a page boundary case
> - fix kvm_arch_vcpu_ioctl_run() should return -EINTR in case ioregionfd
> is interrupted
>
> arch/x86/kvm/vmx/vmx.c | 40 +++++-
> arch/x86/kvm/x86.c | 272 +++++++++++++++++++++++++++++++++++++--
> include/linux/kvm_host.h | 10 ++
> virt/kvm/kvm_main.c | 16 ++-
> 4 files changed, 317 insertions(+), 21 deletions(-)
>
> diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> index 47b8357b9751..39db31afd27e 100644
> --- a/arch/x86/kvm/vmx/vmx.c
> +++ b/arch/x86/kvm/vmx/vmx.c
> @@ -5357,19 +5357,51 @@ static int handle_ept_violation(struct kvm_vcpu *vcpu)
> return kvm_mmu_page_fault(vcpu, gpa, error_code, NULL, 0);
> }
>
> +#ifdef CONFIG_KVM_IOREGION
> +static int complete_ioregion_fast_mmio(struct kvm_vcpu *vcpu)
> +{
> + int ret, idx;
> +
> + idx = srcu_read_lock(&vcpu->kvm->srcu);
> + ret = kvm_io_bus_write(vcpu, KVM_FAST_MMIO_BUS,
> + vcpu->ioregion_ctx.addr, 0, NULL);
> + if (ret) {
> + ret = kvm_mmu_page_fault(vcpu, vcpu->ioregion_ctx.addr,
> + PFERR_RSVD_MASK, NULL, 0);
> + srcu_read_unlock(&vcpu->kvm->srcu, idx);
> + return ret;
> + }
> +
> + srcu_read_unlock(&vcpu->kvm->srcu, idx);
> + return kvm_skip_emulated_instruction(vcpu);
> +}
> +#endif
> +
> static int handle_ept_misconfig(struct kvm_vcpu *vcpu)
> {
> gpa_t gpa;
> + int ret;
>
> /*
> * A nested guest cannot optimize MMIO vmexits, because we have an
> * nGPA here instead of the required GPA.
> */
> gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS);
> - if (!is_guest_mode(vcpu) &&
> - !kvm_io_bus_write(vcpu, KVM_FAST_MMIO_BUS, gpa, 0, NULL)) {
> - trace_kvm_fast_mmio(gpa);
> - return kvm_skip_emulated_instruction(vcpu);
> + if (!is_guest_mode(vcpu)) {
> + ret = kvm_io_bus_write(vcpu, KVM_FAST_MMIO_BUS, gpa, 0, NULL);
> + if (!ret) {
> + trace_kvm_fast_mmio(gpa);
> + return kvm_skip_emulated_instruction(vcpu);
> + }
> +
> +#ifdef CONFIG_KVM_IOREGION
> + if (unlikely(vcpu->ioregion_ctx.is_interrupted && ret == -EINTR)) {
So the question still, EINTR looks wrong which means the syscall can't
be restarted. Not that the syscal doesn't mean KVM_RUN but actually the
kernel_read|write() you want to do with the ioregion fd.
Also do we need to treat differently for EINTR and ERESTARTSYS since
EINTR means the kernel_read()|write() can't be resumed.
Thanks
next prev parent reply other threads:[~2021-03-09 5:52 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-02-21 12:04 [RFC v3 0/5] Introduce MMIO/PIO dispatch file descriptors (ioregionfd) Elena Afanasova
2021-02-21 12:04 ` [RFC v3 1/5] KVM: add initial support for KVM_SET_IOREGION Elena Afanasova
2021-02-24 10:06 ` Stefan Hajnoczi
2021-03-05 13:09 ` Cornelia Huck
2021-03-09 5:26 ` Jason Wang
2021-03-22 9:57 ` Stefan Hajnoczi
2021-02-21 12:04 ` [RFC v3 2/5] KVM: x86: add support for ioregionfd signal handling Elena Afanasova
2021-02-24 10:42 ` Stefan Hajnoczi
2021-03-09 5:51 ` Jason Wang [this message]
2021-03-17 14:19 ` Elena Afanasova
2021-03-26 6:00 ` Jason Wang
2021-02-21 12:04 ` [RFC v3 3/5] KVM: implement wire protocol Elena Afanasova
2021-02-24 11:02 ` Stefan Hajnoczi
2021-03-09 6:19 ` Jason Wang
2021-03-17 13:08 ` Elena Afanasova
2021-03-26 6:21 ` Jason Wang
2021-03-29 16:17 ` Stefan Hajnoczi
2021-02-21 12:04 ` [RFC v3 4/5] KVM: add ioregionfd context Elena Afanasova
2021-02-24 11:27 ` Stefan Hajnoczi
2021-03-09 7:54 ` Jason Wang
2021-03-09 8:01 ` Paolo Bonzini
2021-03-10 13:20 ` Elena Afanasova
2021-03-10 14:11 ` Paolo Bonzini
2021-03-10 16:41 ` Elena Afanasova
[not found] ` <6ff79d0b-3b6a-73d3-ffbd-e4af9758735f@redhat.com>
2021-03-17 10:46 ` Elena Afanasova
2021-03-26 6:47 ` Jason Wang
2021-02-21 12:04 ` [RFC v3 5/5] KVM: enforce NR_IOBUS_DEVS limit if kmemcg is disabled Elena Afanasova
2021-02-21 17:06 ` [RFC v3 0/5] Introduce MMIO/PIO dispatch file descriptors (ioregionfd) Paolo Bonzini
2021-02-22 16:40 ` Elena Afanasova
2021-02-24 11:34 ` Stefan Hajnoczi
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=e276b54a-b2c0-c12e-fdae-22f54824ee6f@redhat.com \
--to=jasowang@redhat.com \
--cc=cohuck@redhat.com \
--cc=eafanasova@gmail.com \
--cc=elena.ufimtseva@oracle.com \
--cc=jag.raman@oracle.com \
--cc=john.levon@nutanix.com \
--cc=kvm@vger.kernel.org \
--cc=mst@redhat.com \
--cc=pbonzini@redhat.com \
--cc=stefanha@redhat.com \
/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).