Re: [PATCH v5 03/15] IOMMU/x86: support freeing of pagetables

From: "Roger Pau Monné" <roger.pau@citrix.com>
To: Jan Beulich <jbeulich@suse.com>
Cc: "xen-devel@lists.xenproject.org" <xen-devel@lists.xenproject.org>,
	Andrew Cooper <andrew.cooper3@citrix.com>,
	Paul Durrant <paul@xen.org>
Subject: Re: [PATCH v5 03/15] IOMMU/x86: support freeing of pagetables
Date: Tue, 31 May 2022 18:25:49 +0200	[thread overview]
Message-ID: <YpZBjVxRdJOzJzZx@Air-de-Roger> (raw)
In-Reply-To: <614413d8-5043-f0e3-929b-f161fa89bb35@suse.com>

On Fri, May 27, 2022 at 01:13:09PM +0200, Jan Beulich wrote:
> For vendor specific code to support superpages we need to be able to
> deal with a superpage mapping replacing an intermediate page table (or
> hierarchy thereof). Consequently an iommu_alloc_pgtable() counterpart is
> needed to free individual page tables while a domain is still alive.
> Since the freeing needs to be deferred until after a suitable IOTLB
> flush was performed, released page tables get queued for processing by a
> tasklet.
> 
> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> ---
> I was considering whether to use a softirq-tasklet instead. This would
> have the benefit of avoiding extra scheduling operations, but come with
> the risk of the freeing happening prematurely because of a
> process_pending_softirqs() somewhere.
> ---
> v5: Fix CPU_UP_PREPARE for BIGMEM. Schedule tasklet in CPU_DOWN_FAILED
>     when list is not empty. Skip all processing in CPU_DEAD when list is
>     empty.
> v4: Change type of iommu_queue_free_pgtable()'s 1st parameter. Re-base.
> v3: Call process_pending_softirqs() from free_queued_pgtables().
> 
> --- a/xen/arch/x86/include/asm/iommu.h
> +++ b/xen/arch/x86/include/asm/iommu.h
> @@ -147,6 +147,7 @@ void iommu_free_domid(domid_t domid, uns
>  int __must_check iommu_free_pgtables(struct domain *d);
>  struct domain_iommu;
>  struct page_info *__must_check iommu_alloc_pgtable(struct domain_iommu *hd);
> +void iommu_queue_free_pgtable(struct domain_iommu *hd, struct page_info *pg);
>  
>  #endif /* !__ARCH_X86_IOMMU_H__ */
>  /*
> --- a/xen/drivers/passthrough/x86/iommu.c
> +++ b/xen/drivers/passthrough/x86/iommu.c
> @@ -12,6 +12,7 @@
>   * this program; If not, see <http://www.gnu.org/licenses/>.
>   */
>  
> +#include <xen/cpu.h>
>  #include <xen/sched.h>
>  #include <xen/iocap.h>
>  #include <xen/iommu.h>
> @@ -566,6 +567,98 @@ struct page_info *iommu_alloc_pgtable(st
>      return pg;
>  }
>  
> +/*
> + * Intermediate page tables which get replaced by large pages may only be
> + * freed after a suitable IOTLB flush. Hence such pages get queued on a
> + * per-CPU list, with a per-CPU tasklet processing the list on the assumption
> + * that the necessary IOTLB flush will have occurred by the time tasklets get
> + * to run. (List and tasklet being per-CPU has the benefit of accesses not
> + * requiring any locking.)
> + */
> +static DEFINE_PER_CPU(struct page_list_head, free_pgt_list);
> +static DEFINE_PER_CPU(struct tasklet, free_pgt_tasklet);
> +
> +static void free_queued_pgtables(void *arg)
> +{
> +    struct page_list_head *list = arg;
> +    struct page_info *pg;
> +    unsigned int done = 0;
> +
> +    while ( (pg = page_list_remove_head(list)) )
> +    {
> +        free_domheap_page(pg);
> +
> +        /* Granularity of checking somewhat arbitrary. */
> +        if ( !(++done & 0x1ff) )
> +             process_pending_softirqs();

Hm, I'm wondering whether we really want to process pending softirqs
here.

Such processing will prevent the watchdog from triggering, which we
likely want in production builds.  OTOH in debug builds we should make
sure that free_queued_pgtables() doesn't take longer than a watchdog
window, or else it's likely to cause issues to guests scheduled on
this same pCPU (and calling process_pending_softirqs() will just mask
it).

Thanks, Roger.