[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v8] x86/mem-sharing: mem-sharing a range of memory



On Wed, Jul 20, 2016 at 7:01 PM, Tamas K Lengyel
<tamas.lengyel@xxxxxxxxxxxx> wrote:
> Currently mem-sharing can be performed on a page-by-page basis from the 
> control
> domain. However, this process is quite wasteful when a range of pages have to
> be deduplicated.
>
> This patch introduces a new mem_sharing memop for range sharing where
> the user doesn't have to separately nominate each page in both the source and
> destination domain, and the looping over all pages happen in the hypervisor.
> This significantly reduces the overhead of sharing a range of memory.
>
> Signed-off-by: Tamas K Lengyel <tamas.lengyel@xxxxxxxxxxxx>
> Acked-by: Wei Liu <wei.liu2@xxxxxxxxxx>
> Reviewed-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
> ---
> Cc: Ian Jackson <ian.jackson@xxxxxxxxxxxxx>
> Cc: George Dunlap <george.dunlap@xxxxxxxxxxxxx>
> Cc: Jan Beulich <jbeulich@xxxxxxxx>
> Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
>
> v8: style fixes and minor adjustments
> ---
>  tools/libxc/include/xenctrl.h        |  15 ++++
>  tools/libxc/xc_memshr.c              |  19 +++++
>  tools/tests/mem-sharing/memshrtool.c |  22 ++++++
>  xen/arch/x86/mm/mem_sharing.c        | 140 
> +++++++++++++++++++++++++++++++++++
>  xen/include/public/memory.h          |  10 ++-
>  5 files changed, 205 insertions(+), 1 deletion(-)
>
> diff --git a/tools/libxc/include/xenctrl.h b/tools/libxc/include/xenctrl.h
> index e904bd5..3782eff 100644
> --- a/tools/libxc/include/xenctrl.h
> +++ b/tools/libxc/include/xenctrl.h
> @@ -2334,6 +2334,21 @@ int xc_memshr_add_to_physmap(xc_interface *xch,
>                      domid_t client_domain,
>                      unsigned long client_gfn);
>
> +/* Allows to deduplicate a range of memory of a client domain. Using
> + * this function is equivalent of calling xc_memshr_nominate_gfn for each gfn
> + * in the two domains followed by xc_memshr_share_gfns.
> + *
> + * May fail with -EINVAL if the source and client domain have different
> + * memory size or if memory sharing is not enabled on either of the domains.
> + * May also fail with -ENOMEM if there isn't enough memory available to store
> + * the sharing metadata before deduplication can happen.
> + */
> +int xc_memshr_range_share(xc_interface *xch,
> +                          domid_t source_domain,
> +                          domid_t client_domain,
> +                          uint64_t start,
> +                          uint64_t end);
> +
>  /* Debug calls: return the number of pages referencing the shared frame 
> backing
>   * the input argument. Should be one or greater.
>   *
> diff --git a/tools/libxc/xc_memshr.c b/tools/libxc/xc_memshr.c
> index deb0aa4..2b871c7 100644
> --- a/tools/libxc/xc_memshr.c
> +++ b/tools/libxc/xc_memshr.c
> @@ -181,6 +181,25 @@ int xc_memshr_add_to_physmap(xc_interface *xch,
>      return xc_memshr_memop(xch, source_domain, &mso);
>  }
>
> +int xc_memshr_range_share(xc_interface *xch,
> +                          domid_t source_domain,
> +                          domid_t client_domain,
> +                          uint64_t start,
> +                          uint64_t end)
> +{
> +    xen_mem_sharing_op_t mso;
> +
> +    memset(&mso, 0, sizeof(mso));
> +
> +    mso.op = XENMEM_sharing_op_range_share;
> +
> +    mso.u.range.client_domain = client_domain;
> +    mso.u.range.start = start;
> +    mso.u.range.end = end;
> +
> +    return xc_memshr_memop(xch, source_domain, &mso);
> +}
> +
>  int xc_memshr_domain_resume(xc_interface *xch,
>                              domid_t domid)
>  {
> diff --git a/tools/tests/mem-sharing/memshrtool.c 
> b/tools/tests/mem-sharing/memshrtool.c
> index 437c7c9..2af6a9e 100644
> --- a/tools/tests/mem-sharing/memshrtool.c
> +++ b/tools/tests/mem-sharing/memshrtool.c
> @@ -24,6 +24,8 @@ static int usage(const char* prog)
>      printf("  nominate <domid> <gfn>  - Nominate a page for sharing.\n");
>      printf("  share <domid> <gfn> <handle> <source> <source-gfn> 
> <source-handle>\n");
>      printf("                          - Share two pages.\n");
> +    printf("  range <source-domid> <destination-domid> <start-gfn> 
> <end-gfn>\n");
> +    printf("                          - Share pages between domains in 
> range.\n");
>      printf("  unshare <domid> <gfn>   - Unshare a page by grabbing a 
> writable map.\n");
>      printf("  add-to-physmap <domid> <gfn> <source> <source-gfn> 
> <source-handle>\n");
>      printf("                          - Populate a page in a domain with a 
> shared page.\n");
> @@ -180,6 +182,26 @@ int main(int argc, const char** argv)
>          }
>          printf("Audit returned %d errors.\n", rc);
>      }
> +    else if( !strcasecmp(cmd, "range") )
> +    {
> +        domid_t sdomid, cdomid;
> +        int rc;
> +        uint64_t start, end;
> +
> +        if ( argc != 6 )
> +            return usage(argv[0]);
>
> +        sdomid = strtol(argv[2], NULL, 0);
> +        cdomid = strtol(argv[3], NULL, 0);
> +        start = strtoul(argv[4], NULL, 0);
> +        end = strtoul(argv[5], NULL, 0);
> +
> +        rc = xc_memshr_range_share(xch, sdomid, cdomid, start, end);
> +        if ( rc < 0 )
> +        {
> +            printf("error executing xc_memshr_range_share: %s\n", 
> strerror(errno));
> +            return rc;
> +        }
> +    }
>      return 0;
>  }
> diff --git a/xen/arch/x86/mm/mem_sharing.c b/xen/arch/x86/mm/mem_sharing.c
> index a522423..329fbd9 100644
> --- a/xen/arch/x86/mm/mem_sharing.c
> +++ b/xen/arch/x86/mm/mem_sharing.c
> @@ -1294,6 +1294,58 @@ int relinquish_shared_pages(struct domain *d)
>      return rc;
>  }
>
> +static int range_share(struct domain *d, struct domain *cd,
> +                       struct mem_sharing_op_range *range)
> +{
> +    int rc = 0;
> +    shr_handle_t sh, ch;
> +    unsigned long start = range->_scratchspace ?: range->start;
> +
> +    while( range->end >= start )
> +    {
> +        /*
> +         * We only break out if we run out of memory as individual pages may
> +         * legitimately be unsharable and we just want to skip over those.
> +         */
> +        rc = mem_sharing_nominate_page(d, start, 0, &sh);
> +        if ( rc == -ENOMEM )
> +            break;
> +
> +        if ( !rc )
> +        {
> +            rc = mem_sharing_nominate_page(cd, start, 0, &ch);
> +            if ( rc == -ENOMEM )
> +                break;
> +
> +            if ( !rc )
> +            {
> +                /* If we get here this should be guaranteed to succeed. */
> +                rc = mem_sharing_share_pages(d, start, sh,
> +                                             cd, start, ch);
> +                ASSERT(!rc);
> +            }
> +        }
> +
> +        /* Check for continuation if it's not the last iteration. */
> +        if ( range->end >= ++start && hypercall_preempt_check() )
> +        {
> +            rc = 1;
> +            break;
> +        }
> +    }
> +
> +    range->_scratchspace = start;
> +
> +    /*
> +     * The last page may fail with -EINVAL, and for range sharing we don't
> +     * care about that.
> +     */
> +    if ( range->end < start && rc == -EINVAL )
> +        rc = 0;
> +
> +    return rc;
> +}
> +
>  int mem_sharing_memop(XEN_GUEST_HANDLE_PARAM(xen_mem_sharing_op_t) arg)
>  {
>      int rc;
> @@ -1468,6 +1520,94 @@ int 
> mem_sharing_memop(XEN_GUEST_HANDLE_PARAM(xen_mem_sharing_op_t) arg)
>          }
>          break;
>
> +        case XENMEM_sharing_op_range_share:
> +        {
> +            unsigned long max_sgfn, max_cgfn;
> +            struct domain *cd;
> +
> +            rc = -EINVAL;
> +            if ( mso.u.range._pad[0] || mso.u.range._pad[1] ||
> +                 mso.u.range._pad[2] )
> +                 goto out;
> +
> +            /*
> +             * We use _scratchscape for the hypercall continuation value.
> +             * Ideally the user sets this to 0 in the beginning but
> +             * there is no good way of enforcing that here, so we just check
> +             * that it's at least in range.
> +             */
> +            if ( mso.u.range._scratchspace &&
> +                 (mso.u.range._scratchspace < mso.u.range.start ||
> +                  mso.u.range._scratchspace > mso.u.range.end) )
> +                goto out;
> +
> +            if ( !mem_sharing_enabled(d) )
> +                goto out;
> +
> +            rc = rcu_lock_live_remote_domain_by_id(mso.u.range.client_domain,
> +                                                   &cd);
> +            if ( rc )
> +                goto out;
> +
> +            /*
> +             * We reuse XENMEM_sharing_op_share XSM check here as this is
> +             * essentially the same concept repeated over multiple pages.
> +             */
> +            rc = xsm_mem_sharing_op(XSM_DM_PRIV, d, cd,
> +                                    XENMEM_sharing_op_share);
> +            if ( rc )
> +            {
> +                rcu_unlock_domain(cd);
> +                goto out;
> +            }
> +
> +            if ( !mem_sharing_enabled(cd) )
> +            {
> +                rcu_unlock_domain(cd);
> +                rc = -EINVAL;
> +                goto out;
> +            }
> +
> +            /*
> +             * Sanity check only, the client should keep the domains paused 
> for
> +             * the duration of this op.
> +             */
> +            if ( !atomic_read(&d->pause_count) ||
> +                 !atomic_read(&cd->pause_count) )
> +            {
> +                rcu_unlock_domain(cd);
> +                rc = -EINVAL;
> +                goto out;
> +            }
> +
> +            max_sgfn = domain_get_maximum_gpfn(d);
> +            max_cgfn = domain_get_maximum_gpfn(cd);
> +
> +            if ( max_sgfn < mso.u.range.start || max_sgfn < mso.u.range.end 
> ||
> +                 max_cgfn < mso.u.range.start || max_cgfn < mso.u.range.end )
> +            {
> +                rcu_unlock_domain(cd);
> +                rc = -EINVAL;
> +                goto out;
> +            }
> +
> +            rc = range_share(d, cd, &mso.u.range);
> +            rcu_unlock_domain(cd);
> +
> +            if ( rc > 0 )
> +            {
> +                if ( __copy_to_guest(arg, &mso, 1) )
> +                    rc = -EFAULT;
> +                else
> +                    rc = 
> hypercall_create_continuation(__HYPERVISOR_memory_op,
> +                                                       "lh", 
> XENMEM_sharing_op,
> +                                                       arg);
> +            }
> +            else
> +                mso.u.range._scratchspace = 0;
> +        }
> +        break;
> +
>          case XENMEM_sharing_op_debug_gfn:
>          {
>              unsigned long gfn = mso.u.debug.u.gfn;
> diff --git a/xen/include/public/memory.h b/xen/include/public/memory.h
> index 29ec571..e0bc018 100644
> --- a/xen/include/public/memory.h
> +++ b/xen/include/public/memory.h
> @@ -465,6 +465,7 @@ DEFINE_XEN_GUEST_HANDLE(xen_mem_access_op_t);
>  #define XENMEM_sharing_op_debug_gref        5
>  #define XENMEM_sharing_op_add_physmap       6
>  #define XENMEM_sharing_op_audit             7
> +#define XENMEM_sharing_op_range_share       8
>
>  #define XENMEM_SHARING_OP_S_HANDLE_INVALID  (-10)
>  #define XENMEM_SHARING_OP_C_HANDLE_INVALID  (-9)
> @@ -500,7 +501,14 @@ struct xen_mem_sharing_op {
>              uint64_aligned_t client_gfn;    /* IN: the client gfn */
>              uint64_aligned_t client_handle; /* IN: handle to the client page 
> */
>              domid_t  client_domain; /* IN: the client domain id */
> -        } share;
> +        } share;
> +        struct mem_sharing_op_range {         /* OP_RANGE_SHARE */
> +            uint64_aligned_t start;          /* IN: start gfn. */
> +            uint64_aligned_t end;            /* IN: end gfn (inclusive) */
> +            uint64_aligned_t _scratchspace;  /* Must be set to 0 */

Tamas,

Why include this "scratchspace" that's not used in the interface,
rather than just doing what the memory operations in
xen/common/memory.c do, and store it in EAX by shifting it over by
MEMOP_EXTENT_SHIFT?  I looked through the history and I see that v1
did something very much like that, but it was changed to using the
scratch space without any explanation.

Having this in the public interface is ugly; and what's worse, it
exposes some of the internal mechanism to the guest.

 -George

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.