Re: [PATCH RFC] vhost: don't use kmap() to log dirty pages

[Jason Wang]

On 2019/5/8 äå12:12, Michael S. Tsirkin wrote:
> On Mon, May 06, 2019 at 10:23:29PM -0400, Jason Wang wrote:
> > Vhost log dirty pages directly to a userspace bitmap through GUP and
> > kmap_atomic() since kernel doesn't have a set_bit_to_user()
> > helper. This will cause issues for the arch that has virtually tagged
> > caches. The way to fix is to keep using userspace virtual address.
> >
> > Fortunately, futex has a cmpxchg to userspace memory helper
> > futex_atomic_cmpxchg_inatomic(). So switch to use it to exchange the
> > userspace bitmap with zero, set the bit and then write it back through
> > put_user().
> >
> > Note: there're archs (few non popular ones) that don't implement
> > futex helper, we can't log dirty pages. We can fix them on top or
> > simply disable LOG_ALL features of vhost.
> Or implement futex_atomic_cmpxchg using kmap if they don't have
> virtually tagged caches.


Yes, this might work.


> > Cc: Christoph Hellwig <hch@xxxxxxxxxxxxx>
> > Cc: James Bottomley <James.Bottomley@xxxxxxxxxxxxxxxxxxxxx>
> > Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
> > Fixes: 3a4d5c94e9593 ("vhost_net: a kernel-level virtio server")
> > Signed-off-by: Jason Wang <jasowang@xxxxxxxxxx>
> > ---
> >   drivers/vhost/vhost.c | 27 +++++++++++++++------------
> >   1 file changed, 15 insertions(+), 12 deletions(-)
> >
> > diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
> > index 351af88..9c94c41 100644
> > --- a/drivers/vhost/vhost.c
> > +++ b/drivers/vhost/vhost.c
> > @@ -31,6 +31,7 @@
> >   #include <linux/sched/signal.h>
> >   #include <linux/interval_tree_generic.h>
> >   #include <linux/nospec.h>
> > +#include <asm/futex.h>
> >   
> >   #include "vhost.h"
> >   
> > @@ -1692,25 +1693,27 @@ long vhost_dev_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *argp)
> >   }
> >   EXPORT_SYMBOL_GPL(vhost_dev_ioctl);
> >   
> > -/* TODO: This is really inefficient.  We need something like get_user()
> > - * (instruction directly accesses the data, with an exception table entry
> > - * returning -EFAULT). See Documentation/x86/exception-tables.txt.
> > - */
> > -static int set_bit_to_user(int nr, void __user *addr)
> > +static int set_bit_to_user(int nr, u32 __user *addr)
> >   {
> >   	unsigned long log = (unsigned long)addr;
> >   	struct page *page;
> > -	void *base;
> > -	int bit = nr + (log % PAGE_SIZE) * 8;
> > +	u32 old_log;
> >   	int r;
> >   
> >   	r = get_user_pages_fast(log, 1, 1, &page);
> >   	if (r < 0)
> >   		return r;
> >   	BUG_ON(r != 1);
> > -	base = kmap_atomic(page);
> > -	set_bit(bit, base);
> > -	kunmap_atomic(base);
> > +
> > +	r = futex_atomic_cmpxchg_inatomic(&old_log, addr, 0, 0);
> > +	if (r < 0)
> > +		return r;
> So I think this is a great idea!
>
> However one issue here is that futex_atomic_cmpxchg_inatomic will fail if the
> page is swapped out. I suspect we need a variant that blocks the thread
> instead.


I guess not since the patch still try to pin the page before.

Thanks


> > +
> > +	old_log |= 1 << nr;
> > +	r = put_user(old_log, addr);
> > +	if (r < 0)
> > +		return r;
> > +
> >   	set_page_dirty_lock(page);
> >   	put_page(page);
> >   	return 0;
> > @@ -1727,8 +1730,8 @@ static int log_write(void __user *log_base,
> >   	write_length += write_address % VHOST_PAGE_SIZE;
> >   	for (;;) {
> >   		u64 base = (u64)(unsigned long)log_base;
> > -		u64 log = base + write_page / 8;
> > -		int bit = write_page % 8;
> > +		u64 log = base + write_page / 32;
> > +		int bit = write_page % 32;
> >   		if ((u64)(unsigned long)log != log)
> >   			return -EFAULT;
> >   		r = set_bit_to_user(bit, (void __user *)(unsigned long)log);
> > --
> > 1.8.3.1