Liu Hui
2008-Dec-04 14:08 UTC
[PATCH]Fix for map_private_extent_buffer to handle extent which is not page alignment
Hi,
Now, BTRFS only supports page alignment extent. But in
map_private_extent(), there are already some codes to handle the
extent which is not page alignment. For example,
map_private_extent_buffer also wants to handle a extent which is 7K
length properly. The problem is the codes are not enough, e.g. in the
following figure, map_private_extent_buffer can handle extent1, but it
can''t take the right behavior to handle extent2. This patch fixs it.
PS:Please use the mono font to view the ASCII figures, thanks.
|-extent1-|
+-----++-----+
|page1||page2|
+-----++-----+
|-extent2-|
+-----++-----+
|page1||page2|
+-----++-----+
--
Thanks & Best Regards
Liu Hui
--
diff --git a/extent_io.c b/extent_io.c
index c3dfe2a..cac47d1 100644
--- a/extent_io.c
+++ b/extent_io.c
@@ -3426,9 +3426,11 @@ int map_private_extent_buffer(struct
extent_buffer *eb, unsigned long start,
char *kaddr;
struct page *p;
size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+ size_t last_offset;
unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
unsigned long end_i = (start_offset + start + min_len - 1) >>
PAGE_CACHE_SHIFT;
+ unsigned long last_i = (start_offset + eb->len - 1) >>
PAGE_CACHE_SHIFT;
if (i != end_i)
return -EINVAL;
@@ -3440,8 +3442,16 @@ int map_private_extent_buffer(struct
extent_buffer *eb, unsigned long start,
offset = 0;
*map_start = ((u64)i << PAGE_CACHE_SHIFT) - start_offset;
}
+
+ if (i == last_i)
+ last_offset = ((u64)(i + 1) << PAGE_CACHE_SHIFT) - start_offset
+ - eb->len;
+ else
+ last_offset = 0;
+
if (start + min_len > eb->len) {
-printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n",
eb->start, eb->len, start, min_len);
+ printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n",
+ eb->start, eb->len, start, min_len);
WARN_ON(1);
}
@@ -3449,7 +3459,7 @@ printk("bad mapping eb start %Lu len %lu, wanted
%lu %lu\n", eb->start, eb->len,
kaddr = kmap_atomic(p, km);
*token = kaddr;
*map = kaddr + offset;
- *map_len = PAGE_CACHE_SIZE - offset;
+ *map_len = PAGE_CACHE_SIZE - offset - last_offset;
return 0;
}
EXPORT_SYMBOL(map_private_extent_buffer);
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Liu Hui
2008-Dec-04 14:21 UTC
Re: [PATCH]Fix for map_private_extent_buffer to handle extent which is not page alignment
There is a little problem with the first ASCII figure, here is the new one. |-extent1-| +-----++-----+ |page1||page2| +-----++-----+ Thanks, Liu Hui 2008/12/4 Liu Hui <onlyflyer@gmail.com>:> Hi, > Now, BTRFS only supports page alignment extent. But in > map_private_extent(), there are already some codes to handle the > extent which is not page alignment. For example, > map_private_extent_buffer also wants to handle a extent which is 7K > length properly. The problem is the codes are not enough, e.g. in the > following figure, map_private_extent_buffer can handle extent1, but it > can''t take the right behavior to handle extent2. This patch fixs it. > > PS:Please use the mono font to view the ASCII figures, thanks. > > |-extent1-| > +-----++-----+ > |page1||page2| > +-----++-----+ > > |-extent2-| > +-----++-----+ > |page1||page2| > +-----++-----+ > > -- > Thanks & Best Regards > Liu Hui > -- > > diff --git a/extent_io.c b/extent_io.c > index c3dfe2a..cac47d1 100644 > --- a/extent_io.c > +++ b/extent_io.c > @@ -3426,9 +3426,11 @@ int map_private_extent_buffer(struct > extent_buffer *eb, unsigned long start, > char *kaddr; > struct page *p; > size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); > + size_t last_offset; > unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; > unsigned long end_i = (start_offset + start + min_len - 1) >> > PAGE_CACHE_SHIFT; > + unsigned long last_i = (start_offset + eb->len - 1) >> PAGE_CACHE_SHIFT; > > if (i != end_i) > return -EINVAL; > @@ -3440,8 +3442,16 @@ int map_private_extent_buffer(struct > extent_buffer *eb, unsigned long start, > offset = 0; > *map_start = ((u64)i << PAGE_CACHE_SHIFT) - start_offset; > } > + > + if (i == last_i) > + last_offset = ((u64)(i + 1) << PAGE_CACHE_SHIFT) - start_offset > + - eb->len; > + else > + last_offset = 0; > + > if (start + min_len > eb->len) { > -printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", > eb->start, eb->len, start, min_len); > + printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", > + eb->start, eb->len, start, min_len); > WARN_ON(1); > } > > @@ -3449,7 +3459,7 @@ printk("bad mapping eb start %Lu len %lu, wanted > %lu %lu\n", eb->start, eb->len, > kaddr = kmap_atomic(p, km); > *token = kaddr; > *map = kaddr + offset; > - *map_len = PAGE_CACHE_SIZE - offset; > + *map_len = PAGE_CACHE_SIZE - offset - last_offset; > return 0; > } > EXPORT_SYMBOL(map_private_extent_buffer); >-- Thanks & Best Regards Liu Hui -- -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Liu Hui
2008-Dec-04 14:28 UTC
Re: [PATCH]Fix for map_private_extent_buffer to handle extent which is not page alignment
Hi,
I don''t know why GMAIL mix my ASCII figure, the figure for extent1 is
still not correct. Here is the newest one, hope GMAIL works.
|-extent1-|
+-----++-----+
|page1||page2|
+-----++-----+
Thanks
2008/12/4 Liu Hui <onlyflyer@gmail.com>:> There is a little problem with the first ASCII figure, here is the new one.
> |-extent1-|
> +-----++-----+
> |page1||page2|
> +-----++-----+
>
> Thanks,
> Liu Hui
>
> 2008/12/4 Liu Hui <onlyflyer@gmail.com>:
>> Hi,
>> Now, BTRFS only supports page alignment extent. But in
>> map_private_extent(), there are already some codes to handle the
>> extent which is not page alignment. For example,
>> map_private_extent_buffer also wants to handle a extent which is 7K
>> length properly. The problem is the codes are not enough, e.g. in the
>> following figure, map_private_extent_buffer can handle extent1, but it
>> can''t take the right behavior to handle extent2. This patch
fixs it.
>>
>> PS:Please use the mono font to view the ASCII figures, thanks.
>>
>> |-extent1-|
>> +-----++-----+
>> |page1||page2|
>> +-----++-----+
>>
>> |-extent2-|
>> +-----++-----+
>> |page1||page2|
>> +-----++-----+
>>
>> --
>> Thanks & Best Regards
>> Liu Hui
>> --
>>
>> diff --git a/extent_io.c b/extent_io.c
>> index c3dfe2a..cac47d1 100644
>> --- a/extent_io.c
>> +++ b/extent_io.c
>> @@ -3426,9 +3426,11 @@ int map_private_extent_buffer(struct
>> extent_buffer *eb, unsigned long start,
>> char *kaddr;
>> struct page *p;
>> size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE -
1);
>> + size_t last_offset;
>> unsigned long i = (start_offset + start) >>
PAGE_CACHE_SHIFT;
>> unsigned long end_i = (start_offset + start + min_len - 1)
>>
>> PAGE_CACHE_SHIFT;
>> + unsigned long last_i = (start_offset + eb->len - 1) >>
PAGE_CACHE_SHIFT;
>>
>> if (i != end_i)
>> return -EINVAL;
>> @@ -3440,8 +3442,16 @@ int map_private_extent_buffer(struct
>> extent_buffer *eb, unsigned long start,
>> offset = 0;
>> *map_start = ((u64)i << PAGE_CACHE_SHIFT) -
start_offset;
>> }
>> +
>> + if (i == last_i)
>> + last_offset = ((u64)(i + 1) << PAGE_CACHE_SHIFT)
- start_offset
>> + - eb->len;
>> + else
>> + last_offset = 0;
>> +
>> if (start + min_len > eb->len) {
>> -printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n",
>> eb->start, eb->len, start, min_len);
>> + printk("bad mapping eb start %Lu len %lu, wanted
%lu %lu\n",
>> + eb->start, eb->len, start, min_len);
>> WARN_ON(1);
>> }
>>
>> @@ -3449,7 +3459,7 @@ printk("bad mapping eb start %Lu len %lu,
wanted
>> %lu %lu\n", eb->start, eb->len,
>> kaddr = kmap_atomic(p, km);
>> *token = kaddr;
>> *map = kaddr + offset;
>> - *map_len = PAGE_CACHE_SIZE - offset;
>> + *map_len = PAGE_CACHE_SIZE - offset - last_offset;
>> return 0;
>> }
>> EXPORT_SYMBOL(map_private_extent_buffer);
>>
>
> --
> Thanks & Best Regards
> Liu Hui
> --
>
--
Thanks & Best Regards
Liu Hui
--
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html