Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1560230

Re: [PATCH] lib: bitmap: introduce bitmap_find_next_zero_area_and_size

From Jaewon Kim <jaewon31.kim@samsung.com>
Newsgroups linux.kernel
Subject Re: [PATCH] lib: bitmap: introduce bitmap_find_next_zero_area_and_size
Date 2017-01-17 04:30 +0100
Message-ID <t0sD7-1yy-1@gated-at.bofh.it> (permalink)
References <sSuVr-82Y-5@gated-at.bofh.it> <sSuVr-82Y-3@gated-at.bofh.it> <t0sD7-1yy-3@gated-at.bofh.it>
Organization linux.* mail to news gateway

Show all headers | View raw



On 2017년 01월 15일 16:17, Yury Norov wrote:
> Hi Jaewon,
>
> with all comments above, some of my concerns.
>
> On Mon, Dec 26, 2016 at 01:18:11PM +0900, Jaewon Kim wrote:
>> There was no bitmap API which returns both next zero index and size of zeros
>> from that index.
> Yes, there is. Most probably because this function is not needed.
> Typical usecase is looking for the area of N free bits, were caller
> knows N, and doesn't care of free areas smaller than N. There is 
> bitmap_find_next_zero_area() for exactly that.
Hi Yuri
Thank you for comment.
I did not mean finding free area but wanted to know its size.
So bitmap_find_next_zero_area is not what I wanted.
I will not submit this patch though.
>
>> This is helpful to look fragmentation. This is an test code to look size of zeros.
>> Test result is '10+9+994=>1013 found of total: 1024'
>>
>> unsigned long search_idx, found_idx, nr_found_tot;
>> unsigned long bitmap_max;
>> unsigned int nr_found;
>> unsigned long *bitmap;
>>
>> search_idx = nr_found_tot = 0;
>> bitmap_max = 1024;
>> bitmap = kzalloc(BITS_TO_LONGS(bitmap_max) * sizeof(long),
>> 		 GFP_KERNEL);
>>
>> /* test bitmap_set offset, count */
>> bitmap_set(bitmap, 10, 1);
>> bitmap_set(bitmap, 20, 10);
>>
>> for (;;) {
>> 	found_idx = bitmap_find_next_zero_area_and_size(bitmap,
>> 				bitmap_max, search_idx, &nr_found);
>> 	if (found_idx >= bitmap_max)
>> 		break;
>> 	if (nr_found_tot == 0)
>> 		printk("%u", nr_found);
>> 	else
>> 		printk("+%u", nr_found);
>> 	nr_found_tot += nr_found;
>> 	search_idx = found_idx + nr_found;
>> }
>> printk("=>%lu found of total: %lu\n", nr_found_tot, bitmap_max);
>>
>> Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
> This usecase is problematic in real world. Consider 1-byte bitmap
> '01010101'. To store fragmentation information for further analysis,
> you have to allocate 4 pairs of address and size. On 64-bit machine
> it's 64 bytes of additional memory. Brief grepping of kernel sources
> shows that no one does it. Correct me if I missed something.
Sorry but I did not understand for "you have to allocate 4 pairs of address and size"
I used just local variables.
>
> If you still think this API is useful, you'd walk over kernel
> and find bins of code that will become better with your function,
> and send the patch that adds the use of your function there. Probable
> candidates for search are bitmap_find_next_zero_area() and find_next_bit()
> functions.
>
> If the only suitable place for new function is your example below, I
> think it's better not to introduce new API and reconsider your
> implementation instead.
>
> Yury.
>
>
>

Back to linux.kernel | Previous | Next | Find similar | Unroll thread


Thread

Re: [PATCH] lib: bitmap: introduce bitmap_find_next_zero_area_and_size Jaewon Kim <jaewon31.kim@samsung.com> - 2017-01-17 04:30 +0100

csiph-web