Help with filing a [maybe] ZFS/mmap bug.

[Andriy Gapon]

on 25/11/2013 20:38 Harald Schmalzbauer said the following:
>  Bezüglich Andriy Gapon's Nachricht vom 18.11.2013 10:38 (localtime):
>> ...
>>>> Here is a patch (for head) that should fix the described above issue:
>>>>
>>>> diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c
>>>> b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c
>>>> index 2e2cbd6..4fcd571 100644
>>>> --- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c
>>>> +++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c
>>>> @@ -328,6 +328,20 @@ page_busy(vnode_t *vp, int64_t start, int64_t off, int64_t
>>>> nbytes)
>>>> {
>>>> 	vm_object_t obj;
>>>> 	vm_page_t pp;
>>>> +	int64_t end;
>>>> +
>>>> +	/*
>>>> +	 * At present vm_page_clear_dirty extends the cleared range to DEV_BSIZE
>>>> +	 * aligned boundaries, if the range is not aligned.  As a result a
>>>> +	 * DEV_BSIZE subrange with partially dirty data may get marked as clean.
>>>> +	 * It may happen that all DEV_BSIZE subranges are marked clean and thus
>>>> +	 * the whole page would be considred clean despite have some dirty data.
>>>> +	 * For this reason we should shrink the range to DEV_BSIZE aligned
>>>> +	 * boundaries before calling vm_page_clear_dirty.
>>>> +	 */
>>>> +	end = rounddown2(off + nbytes, DEV_BSIZE);
>>>> +	off = roundup2(off, DEV_BSIZE);
>>>> +	nbytes = end - off;
>>>>
>>>> 	obj = vp->v_object;
>>>> 	zfs_vmobject_assert_wlocked(obj);
>>>> @@ -362,7 +376,8 @@ page_busy(vnode_t *vp, int64_t start, int64_t off, int64_t
>>>> nbytes)
>>>> 			ASSERT3U(pp->valid, ==, VM_PAGE_BITS_ALL);
>>>> 			vm_object_pip_add(obj, 1);
>>>> 			pmap_remove_write(pp);
>>>> -			vm_page_clear_dirty(pp, off, nbytes);
>>>> +			if (nbytes != 0)
>>>> +				vm_page_clear_dirty(pp, off, nbytes);
>>>> 		}
>>>> 		break;
>>>> 	}
>>>
>>> 9.2 does not seem to have a rounddown2() macro.
>> Thanks for the heads-up!
>> You could use a plain rounddown() or just 'x & ~(DEV_BSIZE - 1)'.
> 
> Hello,
> 
> saw that this fix made it int 9-stable:
> http://svnweb.freebsd.org/base/stable/9/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c?r1=255517&r2=258555
> 
> But it uses rounddown2(), so I guess it's not workit atm, right?

Yeah, sorry about this.
I had a headsup and made a note to myself about it, but then got distracted at
the actual mfc time.
Should be fixed now.

-- 
Andriy Gapon