public inbox for linux-btrfs@vger.kernel.org
 help / color / mirror / Atom feed
From: Qu Wenruo <quwenruo.btrfs@gmx.com>
To: dsterba@suse.cz, Qu Wenruo <wqu@suse.com>, linux-btrfs@vger.kernel.org
Subject: Re: [PATCH v2 07/17] btrfs: make rbio_add_io_page() subpage compatible
Date: Thu, 14 Apr 2022 18:59:58 +0800	[thread overview]
Message-ID: <dce41e95-a11f-3897-bd51-51a10c0da799@gmx.com> (raw)
In-Reply-To: <5b190bf5-892c-0856-e623-f6f716985b28@gmx.com>



On 2022/4/14 08:43, Qu Wenruo wrote:
> 
> 
> On 2022/4/14 07:28, Qu Wenruo wrote:
>>
>>
>> On 2022/4/14 03:14, David Sterba wrote:
>>> There's an assertion failure reported by btrfs/023
>>>
>>> On Tue, Apr 12, 2022 at 05:32:57PM +0800, Qu Wenruo wrote:
>>>> +static int rbio_add_io_sector(struct btrfs_raid_bio *rbio,
>>>> +                  struct bio_list *bio_list,
>>>> +                  struct sector_ptr *sector,
>>>> +                  unsigned int stripe_nr,
>>>> +                  unsigned int sector_nr,
>>>> +                  unsigned long bio_max_len, unsigned int opf)
>>>>   {
>>>> +    const u32 sectorsize = rbio->bioc->fs_info->sectorsize;
>>>>       struct bio *last = bio_list->tail;
>>>>       int ret;
>>>>       struct bio *bio;
>>>>       struct btrfs_io_stripe *stripe;
>>>>       u64 disk_start;
>>>>
>>>> +    /*
>>>> +     * NOTE: here stripe_nr has taken device replace into 
>>>> consideration,
>>>> +     * thus it can be larger than rbio->real_stripe.
>>>> +     * So here we check against bioc->num_stripes, not 
>>>> rbio->real_stripes.
>>>> +     */
>>>> +    ASSERT(stripe_nr >= 0 && stripe_nr < rbio->bioc->num_stripes);
>>>> +    ASSERT(sector_nr >= 0 && sector_nr < rbio->stripe_nsectors);
>>>> +    ASSERT(sector->page);
>>>
>>> This one ^^^
>>
>> Failed to reproduce here, both x86_64 and aarch64 survived over 64 loops
>> of btrfs/023.
>>
>> Although that's withy my github branch. Let me check the current branch
>> to see what's wrong.
> 
> It's a rebase error.
> 
> At least one patch has incorrect diff snippet.
> 
> For the patch "btrfs: raid56: introduce btrfs_raid_bio::stripe_sectors":
> 
> For alloc_rbio(), in my branch and patch (v2) submitted to the mailing 
> list:
> 
> @@ -978,6 +1014,7 @@ static struct btrfs_raid_bio *alloc_rbio(struct 
> btrfs_fs_info *fs_info,
>          rbio = kzalloc(sizeof(*rbio) +
>                         sizeof(*rbio->stripe_pages) * num_pages +
>                         sizeof(*rbio->bio_pages) * num_pages +
> +                      sizeof(*rbio->stripe_sectors) * num_sectors +

The more I check this part of the existing code, the worse it looks to me.

We're really dancing on the edge, it's just a problem of time for us to 
get a cut.

Shouldn't we go the regular way, kzalloc the structure only, then alloc 
needed pages/bitmaps?

Thanks,
Qu
>                         sizeof(*rbio->finish_pointers) * real_stripes +
>                         sizeof(*rbio->dbitmap) * 
> BITS_TO_LONGS(stripe_nsectors) +
>                         sizeof(*rbio->finish_pbitmap) *
> @@ -1015,6 +1052,7 @@ static struct btrfs_raid_bio *alloc_rbio(struct 
> btrfs_fs_info *fs_info,
>          } while (0)
>          CONSUME_ALLOC(rbio->stripe_pages, num_pages);
>          CONSUME_ALLOC(rbio->bio_pages, num_pages);
> +       CONSUME_ALLOC(rbio->stripe_sectors, num_sectors);
>          CONSUME_ALLOC(rbio->finish_pointers, real_stripes);
>          CONSUME_ALLOC(rbio->dbitmap, BITS_TO_LONGS(stripe_nsectors));
>          CONSUME_ALLOC(rbio->finish_pbitmap, 
> BITS_TO_LONGS(stripe_nsectors));
> 
> It has the correct extra space for stripe_sectors.
> 
> But in the for-next branch, it looks like this:
> @@ -1014,6 +1050,7 @@ static struct btrfs_raid_bio *alloc_rbio(struct 
> btrfs_fs_info *fs_info,
>          } while (0)
>          CONSUME_ALLOC(rbio->stripe_pages, num_pages);
>          CONSUME_ALLOC(rbio->bio_pages, num_pages);
> +       CONSUME_ALLOC(rbio->stripe_sectors, num_sectors);
>          CONSUME_ALLOC(rbio->finish_pointers, real_stripes);
>          CONSUME_ALLOC(rbio->dbitmap, BITS_TO_LONGS(stripe_nsectors));
>          CONSUME_ALLOC(rbio->finish_pbitmap, 
> BITS_TO_LONGS(stripe_nsectors));
> 
> Without the extra space allocated for stripe_sectors, no wonder it crashes.
> 
> Mind to provide a branch for me to rebase my patchse upon?
> 
> Thanks,
> Qu
> 
>>
>> Thanks,
>> Qu
>>
>>>
>>> [ 2280.705765] assertion failed: sector->page, in fs/btrfs/raid56.c:1145
>>> [ 2280.707844] ------------[ cut here ]------------
>>> [ 2280.709401] kernel BUG at fs/btrfs/ctree.h:3614!
>>> [ 2280.711084] invalid opcode: 0000 [#1] PREEMPT SMP
>>> [ 2280.712822] CPU: 3 PID: 4084 Comm: kworker/u8:2 Not tainted 
>>> 5.18.0-rc2-default+ #1697
>>> [ 2280.715648] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), 
>>> BIOS rel-1.14.0-0-g155821a-rebuilt.opensuse.org 04/01/2014
>>> [ 2280.719656] Workqueue: btrfs-rmw btrfs_work_helper [btrfs]
>>> [ 2280.721775] RIP: 0010:assertfail.constprop.0+0x18/0x1a [btrfs]
>>> [ 2280.729575] RSP: 0018:ffffad6d071afda0 EFLAGS: 00010246
>>> [ 2280.730844] RAX: 0000000000000039 RBX: 0000000000000000 RCX: 
>>> 0000000000000000
>>> [ 2280.732449] RDX: 0000000000000000 RSI: 0000000000000003 RDI: 
>>> 00000000ffffffff
>>> [ 2280.733992] RBP: ffff8e51d5465000 R08: 0000000000000003 R09: 
>>> 0000000000000002
>>> [ 2280.735535] R10: 0000000000000000 R11: 0000000000000001 R12: 
>>> 0000000000000003
>>> [ 2280.737093] R13: ffff8e51d5465000 R14: ffff8e51d5465d78 R15: 
>>> 0000000000001000
>>> [ 2280.738613] FS:  0000000000000000(0000) GS:ffff8e523dc00000(0000) 
>>> knlGS:0000000000000000
>>> [ 2280.740392] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>>> [ 2280.741794] CR2: 000055f48fe51ab0 CR3: 000000001f805001 CR4: 
>>> 0000000000170ea0
>>> [ 2280.743532] Call Trace:
>>> [ 2280.744136]  <TASK>
>>> [ 2280.744701]  rbio_add_io_sector.cold+0x11/0x33 [btrfs]
>>> [ 2280.745846]  ? _raw_spin_unlock_irq+0x2f/0x50
>>> [ 2280.746782]  raid56_rmw_stripe.isra.0+0x153/0x320 [btrfs]
>>> [ 2280.747965]  btrfs_work_helper+0xd6/0x1d0 [btrfs]
>>> [ 2280.749018]  process_one_work+0x264/0x5f0
>>> [ 2280.749806]  worker_thread+0x52/0x3b0
>>> [ 2280.750523]  ? process_one_work+0x5f0/0x5f0
>>> [ 2280.751395]  kthread+0xea/0x110
>>> [ 2280.752097]  ? kthread_complete_and_exit+0x20/0x20
>>> [ 2280.753112]  ret_from_fork+0x1f/0x30
>>>
>>>
>>>> +
>>>> +    /* We don't yet support subpage, thus pgoff should always be 0 */
>>>> +    ASSERT(sector->pgoff == 0);
>>>> +
>>>>       stripe = &rbio->bioc->stripes[stripe_nr];
>>>> -    disk_start = stripe->physical + (page_index << PAGE_SHIFT);
>>>> +    disk_start = stripe->physical + sector_nr * sectorsize;
>>>>
>>>>       /* if the device is missing, just fail this stripe */
>>>>       if (!stripe->dev->bdev)
>>>> @@ -1156,8 +1226,9 @@ static int rbio_add_io_page(struct 
>>>> btrfs_raid_bio *rbio,
>>>>            */
>>>>           if (last_end == disk_start && !last->bi_status &&
>>>>               last->bi_bdev == stripe->dev->bdev) {
>>>> -            ret = bio_add_page(last, page, PAGE_SIZE, 0);
>>>> -            if (ret == PAGE_SIZE)
>>>> +            ret = bio_add_page(last, sector->page, sectorsize,
>>>> +                       sector->pgoff);
>>>> +            if (ret == sectorsize)
>>>>                   return 0;
>>>>           }
>>>>       }
>>>> @@ -1168,7 +1239,7 @@ static int rbio_add_io_page(struct 
>>>> btrfs_raid_bio *rbio,
>>>>       bio->bi_iter.bi_sector = disk_start >> 9;
>>>>       bio->bi_private = rbio;
>>>>
>>>> -    bio_add_page(bio, page, PAGE_SIZE, 0);
>>>> +    bio_add_page(bio, sector->page, sectorsize, sector->pgoff);
>>>>       bio_list_add(bio_list, bio);
>>>>       return 0;
>>>>   }
>>>> @@ -1265,7 +1336,7 @@ static noinline void finish_rmw(struct 
>>>> btrfs_raid_bio *rbio)
>>>>       void **pointers = rbio->finish_pointers;
>>>>       int nr_data = rbio->nr_data;
>>>>       int stripe;
>>>> -    int pagenr;
>>>> +    int sectornr;
>>>>       bool has_qstripe;
>>>>       struct bio_list bio_list;
>>>>       struct bio *bio;
>>>> @@ -1309,16 +1380,16 @@ static noinline void finish_rmw(struct 
>>>> btrfs_raid_bio *rbio)
>>>>       else
>>>>           clear_bit(RBIO_CACHE_READY_BIT, &rbio->flags);
>>>>
>>>> -    for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) {
>>>> +    for (sectornr = 0; sectornr < rbio->stripe_nsectors; sectornr++) {
>>>>           struct page *p;
>>>>           /* first collect one page from each data stripe */
>>>>           for (stripe = 0; stripe < nr_data; stripe++) {
>>>> -            p = page_in_rbio(rbio, stripe, pagenr, 0);
>>>> +            p = page_in_rbio(rbio, stripe, sectornr, 0);
>>>>               pointers[stripe] = kmap_local_page(p);
>>>>           }
>>>>
>>>>           /* then add the parity stripe */
>>>> -        p = rbio_pstripe_page(rbio, pagenr);
>>>> +        p = rbio_pstripe_page(rbio, sectornr);
>>>>           SetPageUptodate(p);
>>>>           pointers[stripe++] = kmap_local_page(p);
>>>>
>>>> @@ -1328,7 +1399,7 @@ static noinline void finish_rmw(struct 
>>>> btrfs_raid_bio *rbio)
>>>>                * raid6, add the qstripe and call the
>>>>                * library function to fill in our p/q
>>>>                */
>>>> -            p = rbio_qstripe_page(rbio, pagenr);
>>>> +            p = rbio_qstripe_page(rbio, sectornr);
>>>>               SetPageUptodate(p);
>>>>               pointers[stripe++] = kmap_local_page(p);
>>>>
>>>> @@ -1349,19 +1420,19 @@ static noinline void finish_rmw(struct 
>>>> btrfs_raid_bio *rbio)
>>>>        * everything else.
>>>>        */
>>>>       for (stripe = 0; stripe < rbio->real_stripes; stripe++) {
>>>> -        for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) {
>>>> -            struct page *page;
>>>> +        for (sectornr = 0; sectornr < rbio->stripe_nsectors; 
>>>> sectornr++) {
>>>> +            struct sector_ptr *sector;
>>>> +
>>>>               if (stripe < rbio->nr_data) {
>>>> -                page = page_in_rbio(rbio, stripe, pagenr, 1);
>>>> -                if (!page)
>>>> +                sector = sector_in_rbio(rbio, stripe, sectornr, 1);
>>>> +                if (!sector)
>>>>                       continue;
>>>>               } else {
>>>> -                   page = rbio_stripe_page(rbio, stripe, pagenr);
>>>> +                sector = rbio_stripe_sector(rbio, stripe, sectornr);
>>>>               }
>>>>
>>>> -            ret = rbio_add_io_page(rbio, &bio_list,
>>>> -                       page, stripe, pagenr, rbio->stripe_len,
>>>> -                       REQ_OP_WRITE);
>>>> +            ret = rbio_add_io_sector(rbio, &bio_list, sector, stripe,
>>>> +                         sectornr, rbio->stripe_len, REQ_OP_WRITE);
>>>>               if (ret)
>>>>                   goto cleanup;
>>>>           }
>>>> @@ -1374,20 +1445,21 @@ static noinline void finish_rmw(struct 
>>>> btrfs_raid_bio *rbio)
>>>>           if (!bioc->tgtdev_map[stripe])
>>>>               continue;
>>>>
>>>> -        for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) {
>>>> -            struct page *page;
>>>> +        for (sectornr = 0; sectornr < rbio->stripe_nsectors; 
>>>> sectornr++) {
>>>> +            struct sector_ptr *sector;
>>>> +
>>>>               if (stripe < rbio->nr_data) {
>>>> -                page = page_in_rbio(rbio, stripe, pagenr, 1);
>>>> -                if (!page)
>>>> +                sector = sector_in_rbio(rbio, stripe, sectornr, 1);
>>>> +                if (!sector)
>>>>                       continue;
>>>>               } else {
>>>> -                   page = rbio_stripe_page(rbio, stripe, pagenr);
>>>> +                sector = rbio_stripe_sector(rbio, stripe, sectornr);
>>>>               }
>>>>
>>>> -            ret = rbio_add_io_page(rbio, &bio_list, page,
>>>> -                           rbio->bioc->tgtdev_map[stripe],
>>>> -                           pagenr, rbio->stripe_len,
>>>> -                           REQ_OP_WRITE);
>>>> +            ret = rbio_add_io_sector(rbio, &bio_list, sector,
>>>> +                         rbio->bioc->tgtdev_map[stripe],
>>>> +                         sectornr, rbio->stripe_len,
>>>> +                         REQ_OP_WRITE);
>>>>               if (ret)
>>>>                   goto cleanup;
>>>>           }
>>>> @@ -1563,7 +1635,7 @@ static int raid56_rmw_stripe(struct 
>>>> btrfs_raid_bio *rbio)
>>>>       int bios_to_read = 0;
>>>>       struct bio_list bio_list;
>>>>       int ret;
>>>> -    int pagenr;
>>>> +    int sectornr;
>>>>       int stripe;
>>>>       struct bio *bio;
>>>>
>>>> @@ -1581,28 +1653,29 @@ static int raid56_rmw_stripe(struct 
>>>> btrfs_raid_bio *rbio)
>>>>        * stripe
>>>>        */
>>>>       for (stripe = 0; stripe < rbio->nr_data; stripe++) {
>>>> -        for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) {
>>>> -            struct page *page;
>>>> +        for (sectornr = 0; sectornr < rbio->stripe_nsectors; 
>>>> sectornr++) {
>>>> +            struct sector_ptr *sector;
>>>> +
>>>>               /*
>>>> -             * we want to find all the pages missing from
>>>> +             * We want to find all the sectors missing from
>>>>                * the rbio and read them from the disk.  If
>>>> -             * page_in_rbio finds a page in the bio list
>>>> +             * sector_in_rbio() finds a page in the bio list
>>>>                * we don't need to read it off the stripe.
>>>>                */
>>>> -            page = page_in_rbio(rbio, stripe, pagenr, 1);
>>>> -            if (page)
>>>> +            sector = sector_in_rbio(rbio, stripe, sectornr, 1);
>>>> +            if (sector)
>>>>                   continue;
>>>>
>>>> -            page = rbio_stripe_page(rbio, stripe, pagenr);
>>>> +            sector = rbio_stripe_sector(rbio, stripe, sectornr);
>>>>               /*
>>>> -             * the bio cache may have handed us an uptodate
>>>> +             * The bio cache may have handed us an uptodate
>>>>                * page.  If so, be happy and use it
>>>>                */
>>>> -            if (PageUptodate(page))
>>>> +            if (sector->uptodate)
>>>>                   continue;
>>>>
>>>> -            ret = rbio_add_io_page(rbio, &bio_list, page,
>>>> -                       stripe, pagenr, rbio->stripe_len,
>>>> +            ret = rbio_add_io_sector(rbio, &bio_list, sector,
>>>> +                       stripe, sectornr, rbio->stripe_len,
>>>>                          REQ_OP_READ);
>>>>               if (ret)
>>>>                   goto cleanup;
>>>> @@ -2107,7 +2180,7 @@ static int __raid56_parity_recover(struct 
>>>> btrfs_raid_bio *rbio)
>>>>       int bios_to_read = 0;
>>>>       struct bio_list bio_list;
>>>>       int ret;
>>>> -    int pagenr;
>>>> +    int sectornr;
>>>>       int stripe;
>>>>       struct bio *bio;
>>>>
>>>> @@ -2130,21 +2203,20 @@ static int __raid56_parity_recover(struct 
>>>> btrfs_raid_bio *rbio)
>>>>               continue;
>>>>           }
>>>>
>>>> -        for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) {
>>>> -            struct page *p;
>>>> +        for (sectornr = 0; sectornr < rbio->stripe_nsectors; 
>>>> sectornr++) {
>>>> +            struct sector_ptr *sector;
>>>>
>>>>               /*
>>>>                * the rmw code may have already read this
>>>>                * page in
>>>>                */
>>>> -            p = rbio_stripe_page(rbio, stripe, pagenr);
>>>> -            if (PageUptodate(p))
>>>> +            sector = rbio_stripe_sector(rbio, stripe, sectornr);
>>>> +            if (sector->uptodate)
>>>>                   continue;
>>>>
>>>> -            ret = rbio_add_io_page(rbio, &bio_list,
>>>> -                       rbio_stripe_page(rbio, stripe, pagenr),
>>>> -                       stripe, pagenr, rbio->stripe_len,
>>>> -                       REQ_OP_READ);
>>>> +            ret = rbio_add_io_sector(rbio, &bio_list, sector,
>>>> +                         stripe, sectornr,
>>>> +                         rbio->stripe_len, REQ_OP_READ);
>>>>               if (ret < 0)
>>>>                   goto cleanup;
>>>>           }
>>>> @@ -2399,7 +2471,7 @@ static noinline void 
>>>> finish_parity_scrub(struct btrfs_raid_bio *rbio,
>>>>       unsigned long *pbitmap = rbio->finish_pbitmap;
>>>>       int nr_data = rbio->nr_data;
>>>>       int stripe;
>>>> -    int pagenr;
>>>> +    int sectornr;
>>>>       bool has_qstripe;
>>>>       struct page *p_page = NULL;
>>>>       struct page *q_page = NULL;
>>>> @@ -2419,7 +2491,7 @@ static noinline void 
>>>> finish_parity_scrub(struct btrfs_raid_bio *rbio,
>>>>
>>>>       if (bioc->num_tgtdevs && bioc->tgtdev_map[rbio->scrubp]) {
>>>>           is_replace = 1;
>>>> -        bitmap_copy(pbitmap, rbio->dbitmap, rbio->stripe_npages);
>>>> +        bitmap_copy(pbitmap, rbio->dbitmap, rbio->stripe_nsectors);
>>>>       }
>>>>
>>>>       /*
>>>> @@ -2453,12 +2525,12 @@ static noinline void 
>>>> finish_parity_scrub(struct btrfs_raid_bio *rbio,
>>>>       /* Map the parity stripe just once */
>>>>       pointers[nr_data] = kmap_local_page(p_page);
>>>>
>>>> -    for_each_set_bit(pagenr, rbio->dbitmap, rbio->stripe_npages) {
>>>> +    for_each_set_bit(sectornr, rbio->dbitmap, rbio->stripe_nsectors) {
>>>>           struct page *p;
>>>>           void *parity;
>>>>           /* first collect one page from each data stripe */
>>>>           for (stripe = 0; stripe < nr_data; stripe++) {
>>>> -            p = page_in_rbio(rbio, stripe, pagenr, 0);
>>>> +            p = page_in_rbio(rbio, stripe, sectornr, 0);
>>>>               pointers[stripe] = kmap_local_page(p);
>>>>           }
>>>>
>>>> @@ -2473,13 +2545,13 @@ static noinline void 
>>>> finish_parity_scrub(struct btrfs_raid_bio *rbio,
>>>>           }
>>>>
>>>>           /* Check scrubbing parity and repair it */
>>>> -        p = rbio_stripe_page(rbio, rbio->scrubp, pagenr);
>>>> +        p = rbio_stripe_page(rbio, rbio->scrubp, sectornr);
>>>>           parity = kmap_local_page(p);
>>>>           if (memcmp(parity, pointers[rbio->scrubp], PAGE_SIZE))
>>>>               copy_page(parity, pointers[rbio->scrubp]);
>>>>           else
>>>>               /* Parity is right, needn't writeback */
>>>> -            bitmap_clear(rbio->dbitmap, pagenr, 1);
>>>> +            bitmap_clear(rbio->dbitmap, sectornr, 1);
>>>>           kunmap_local(parity);
>>>>
>>>>           for (stripe = nr_data - 1; stripe >= 0; stripe--)
>>>> @@ -2499,12 +2571,13 @@ static noinline void 
>>>> finish_parity_scrub(struct btrfs_raid_bio *rbio,
>>>>        * higher layers (the bio_list in our rbio) and our p/q.  Ignore
>>>>        * everything else.
>>>>        */
>>>> -    for_each_set_bit(pagenr, rbio->dbitmap, rbio->stripe_npages) {
>>>> -        struct page *page;
>>>> +    for_each_set_bit(sectornr, rbio->dbitmap, rbio->stripe_nsectors) {
>>>> +        struct sector_ptr *sector;
>>>>
>>>> -        page = rbio_stripe_page(rbio, rbio->scrubp, pagenr);
>>>> -        ret = rbio_add_io_page(rbio, &bio_list, page, rbio->scrubp,
>>>> -                       pagenr, rbio->stripe_len, REQ_OP_WRITE);
>>>> +        sector = rbio_stripe_sector(rbio, rbio->scrubp, sectornr);
>>>> +        ret = rbio_add_io_sector(rbio, &bio_list, sector, 
>>>> rbio->scrubp,
>>>> +                     sectornr, rbio->stripe_len,
>>>> +                     REQ_OP_WRITE);
>>>>           if (ret)
>>>>               goto cleanup;
>>>>       }
>>>> @@ -2512,13 +2585,13 @@ static noinline void 
>>>> finish_parity_scrub(struct btrfs_raid_bio *rbio,
>>>>       if (!is_replace)
>>>>           goto submit_write;
>>>>
>>>> -    for_each_set_bit(pagenr, pbitmap, rbio->stripe_npages) {
>>>> -        struct page *page;
>>>> +    for_each_set_bit(sectornr, pbitmap, rbio->stripe_nsectors) {
>>>> +        struct sector_ptr *sector;
>>>>
>>>> -        page = rbio_stripe_page(rbio, rbio->scrubp, pagenr);
>>>> -        ret = rbio_add_io_page(rbio, &bio_list, page,
>>>> +        sector = rbio_stripe_sector(rbio, rbio->scrubp, sectornr);
>>>> +        ret = rbio_add_io_sector(rbio, &bio_list, sector,
>>>>                          bioc->tgtdev_map[rbio->scrubp],
>>>> -                       pagenr, rbio->stripe_len, REQ_OP_WRITE);
>>>> +                       sectornr, rbio->stripe_len, REQ_OP_WRITE);
>>>>           if (ret)
>>>>               goto cleanup;
>>>>       }
>>>> @@ -2650,7 +2723,7 @@ static void raid56_parity_scrub_stripe(struct 
>>>> btrfs_raid_bio *rbio)
>>>>       int bios_to_read = 0;
>>>>       struct bio_list bio_list;
>>>>       int ret;
>>>> -    int pagenr;
>>>> +    int sectornr;
>>>>       int stripe;
>>>>       struct bio *bio;
>>>>
>>>> @@ -2666,28 +2739,30 @@ static void 
>>>> raid56_parity_scrub_stripe(struct btrfs_raid_bio *rbio)
>>>>        * stripe
>>>>        */
>>>>       for (stripe = 0; stripe < rbio->real_stripes; stripe++) {
>>>> -        for_each_set_bit(pagenr, rbio->dbitmap, rbio->stripe_npages) {
>>>> -            struct page *page;
>>>> +        for_each_set_bit(sectornr, rbio->dbitmap,
>>>> +                 rbio->stripe_nsectors) {
>>>> +            struct sector_ptr *sector;
>>>>               /*
>>>> -             * we want to find all the pages missing from
>>>> +             * We want to find all the sectors missing from
>>>>                * the rbio and read them from the disk.  If
>>>> -             * page_in_rbio finds a page in the bio list
>>>> +             * sector_in_rbio() finds a sector in the bio list
>>>>                * we don't need to read it off the stripe.
>>>>                */
>>>> -            page = page_in_rbio(rbio, stripe, pagenr, 1);
>>>> -            if (page)
>>>> +            sector = sector_in_rbio(rbio, stripe, sectornr, 1);
>>>> +            if (sector)
>>>>                   continue;
>>>>
>>>> -            page = rbio_stripe_page(rbio, stripe, pagenr);
>>>> +            sector = rbio_stripe_sector(rbio, stripe, sectornr);
>>>>               /*
>>>> -             * the bio cache may have handed us an uptodate
>>>> -             * page.  If so, be happy and use it
>>>> +             * The bio cache may have handed us an uptodate
>>>> +             * sector.  If so, be happy and use it
>>>>                */
>>>> -            if (PageUptodate(page))
>>>> +            if (sector->uptodate)
>>>>                   continue;
>>>>
>>>> -            ret = rbio_add_io_page(rbio, &bio_list, page, stripe,
>>>> -                           pagenr, rbio->stripe_len, REQ_OP_READ);
>>>> +            ret = rbio_add_io_sector(rbio, &bio_list, sector,
>>>> +                         stripe, sectornr,
>>>> +                         rbio->stripe_len, REQ_OP_READ);
>>>>               if (ret)
>>>>                   goto cleanup;
>>>>           }
>>>> -- 
>>>> 2.35.1

  reply	other threads:[~2022-04-14 11:00 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-04-12  9:32 [PATCH v2 00/17] btrfs: add subpage support for RAID56 Qu Wenruo
2022-04-12  9:32 ` [PATCH v2 01/17] btrfs: reduce width for stripe_len from u64 to u32 Qu Wenruo
2022-04-12  9:32 ` [PATCH v2 02/17] btrfs: open-code rbio_nr_pages() Qu Wenruo
2022-04-12  9:32 ` [PATCH v2 03/17] btrfs: make btrfs_raid_bio more compact Qu Wenruo
2022-04-12  9:32 ` [PATCH v2 04/17] btrfs: introduce new cached members for btrfs_raid_bio Qu Wenruo
2022-04-15  5:31   ` Christoph Hellwig
2022-04-15  5:34     ` Qu Wenruo
2022-04-15  5:45       ` Christoph Hellwig
2022-04-12  9:32 ` [PATCH v2 05/17] btrfs: introduce btrfs_raid_bio::stripe_sectors Qu Wenruo
2022-04-12  9:32 ` [PATCH v2 06/17] btrfs: introduce btrfs_raid_bio::bio_sectors Qu Wenruo
2022-04-12  9:32 ` [PATCH v2 07/17] btrfs: make rbio_add_io_page() subpage compatible Qu Wenruo
2022-04-13 19:14   ` David Sterba
2022-04-13 23:28     ` Qu Wenruo
2022-04-14  0:43       ` Qu Wenruo
2022-04-14 10:59         ` Qu Wenruo [this message]
2022-04-14 15:51           ` David Sterba
2022-04-14 22:48             ` Qu Wenruo
2022-04-21 15:44               ` David Sterba
2022-04-14 15:43         ` David Sterba
2022-04-14 17:51           ` David Sterba
2022-04-14 22:28             ` Qu Wenruo
2022-04-21 16:24               ` David Sterba
2022-04-12  9:32 ` [PATCH v2 08/17] btrfs: make finish_parity_scrub() " Qu Wenruo
2022-04-12  9:32 ` [PATCH v2 09/17] btrfs: make __raid_recover_endio_io() subpage compatibable Qu Wenruo
2022-04-12  9:33 ` [PATCH v2 10/17] btrfs: make finish_rmw() subpage compatible Qu Wenruo
2022-04-12  9:33 ` [PATCH v2 11/17] btrfs: open-code rbio_stripe_page_index() Qu Wenruo
2022-04-12  9:33 ` [PATCH v2 12/17] btrfs: make raid56_add_scrub_pages() subpage compatible Qu Wenruo
2022-04-12  9:33 ` [PATCH v2 13/17] btrfs: remove btrfs_raid_bio::bio_pages array Qu Wenruo
2022-04-12  9:33 ` [PATCH v2 14/17] btrfs: make set_bio_pages_uptodate() subpage compatible Qu Wenruo
2022-04-12  9:33 ` [PATCH v2 15/17] btrfs: make steal_rbio() " Qu Wenruo
2022-04-12  9:33 ` [PATCH v2 16/17] btrfs: make alloc_rbio_essential_pages() " Qu Wenruo
2022-04-12  9:33 ` [PATCH v2 17/17] btrfs: enable subpage support for RAID56 Qu Wenruo
2022-04-12 17:42 ` [PATCH v2 00/17] btrfs: add " David Sterba
2022-04-13 14:46   ` David Sterba

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=dce41e95-a11f-3897-bd51-51a10c0da799@gmx.com \
    --to=quwenruo.btrfs@gmx.com \
    --cc=dsterba@suse.cz \
    --cc=linux-btrfs@vger.kernel.org \
    --cc=wqu@suse.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox