From: Qu Wenruo <wqu@suse.com>
Date: Wed, 9 Apr 2025 11:10:40 +0000 (+0200)
Subject: btrfs: raid56: store a physical address in structure sector_ptr
X-Git-Tag: block-6.16-20250606~42^2~114
X-Git-Url: https://git.kernel.dk/?a=commitdiff_plain;h=cd678925e9f20e671fdb67a9e1d358ad445a8cc9;p=linux-block.git

btrfs: raid56: store a physical address in structure sector_ptr

Instead of using a @page + @pg_offset pair inside sector_ptr structure,
use a single physical address instead.

This allows us to grab both the page and offset from a single u64 value.
Although we still need an extra bool value, @has_paddr, to distinguish
if the sector is properly mapped (as the 0 physical address is totally
valid).

This change doesn't change the size of structure sector_ptr, but reduces
the parameters of several functions.

Note: the original idea and patch is from Christoph Hellwig
(https://lore.kernel.org/linux-btrfs/20250409111055.3640328-7-hch@lst.de/)
but the final implementation is different.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
[ Use physical addresses instead to handle highmem. ]
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
---

diff --git a/fs/btrfs/raid56.c b/fs/btrfs/raid56.c
index 6e1d9cdd59c1..09ae1018221e 100644
--- a/fs/btrfs/raid56.c
+++ b/fs/btrfs/raid56.c
@@ -134,14 +134,17 @@ struct btrfs_stripe_hash_table {
 };
 
 /*
- * A bvec like structure to present a sector inside a page.
- *
- * Unlike bvec we don't need bvlen, as it's fixed to sectorsize.
+ * A structure to present a sector inside a page, the length is fixed to
+ * sectorsize;
  */
 struct sector_ptr {
-	struct page *page;
-	unsigned int pgoff:24;
-	unsigned int uptodate:8;
+	/*
+	 * Blocks from the bio list can still be highmem.
+	 * So here we use physical address to present a page and the offset inside it.
+	 */
+	phys_addr_t paddr;
+	bool has_paddr;
+	bool uptodate;
 };
 
 static void rmw_rbio_work(struct work_struct *work);
@@ -233,6 +236,14 @@ int btrfs_alloc_stripe_hash_table(struct btrfs_fs_info *info)
 	return 0;
 }
 
+static void memcpy_sectors(const struct sector_ptr *dst,
+			   const struct sector_ptr *src, u32 blocksize)
+{
+	memcpy_page(phys_to_page(dst->paddr), offset_in_page(dst->paddr),
+		    phys_to_page(src->paddr), offset_in_page(src->paddr),
+		    blocksize);
+}
+
 /*
  * caching an rbio means to copy anything from the
  * bio_sectors array into the stripe_pages array.  We
@@ -253,7 +264,7 @@ static void cache_rbio_pages(struct btrfs_raid_bio *rbio)
 
 	for (i = 0; i < rbio->nr_sectors; i++) {
 		/* Some range not covered by bio (partial write), skip it */
-		if (!rbio->bio_sectors[i].page) {
+		if (!rbio->bio_sectors[i].has_paddr) {
 			/*
 			 * Even if the sector is not covered by bio, if it is
 			 * a data sector it should still be uptodate as it is
@@ -264,12 +275,8 @@ static void cache_rbio_pages(struct btrfs_raid_bio *rbio)
 			continue;
 		}
 
-		ASSERT(rbio->stripe_sectors[i].page);
-		memcpy_page(rbio->stripe_sectors[i].page,
-			    rbio->stripe_sectors[i].pgoff,
-			    rbio->bio_sectors[i].page,
-			    rbio->bio_sectors[i].pgoff,
-			    rbio->bioc->fs_info->sectorsize);
+		memcpy_sectors(&rbio->stripe_sectors[i], &rbio->bio_sectors[i],
+				rbio->bioc->fs_info->sectorsize);
 		rbio->stripe_sectors[i].uptodate = 1;
 	}
 	set_bit(RBIO_CACHE_READY_BIT, &rbio->flags);
@@ -326,8 +333,13 @@ static void index_stripe_sectors(struct btrfs_raid_bio *rbio)
 		int page_index = offset >> PAGE_SHIFT;
 
 		ASSERT(page_index < rbio->nr_pages);
-		rbio->stripe_sectors[i].page = rbio->stripe_pages[page_index];
-		rbio->stripe_sectors[i].pgoff = offset_in_page(offset);
+		if (!rbio->stripe_pages[page_index])
+			continue;
+
+		rbio->stripe_sectors[i].has_paddr = true;
+		rbio->stripe_sectors[i].paddr =
+			page_to_phys(rbio->stripe_pages[page_index]) +
+			offset_in_page(offset);
 	}
 }
 
@@ -962,9 +974,9 @@ static struct sector_ptr *sector_in_rbio(struct btrfs_raid_bio *rbio,
 
 	spin_lock(&rbio->bio_list_lock);
 	sector = &rbio->bio_sectors[index];
-	if (sector->page || bio_list_only) {
+	if (sector->has_paddr || bio_list_only) {
 		/* Don't return sector without a valid page pointer */
-		if (!sector->page)
+		if (!sector->has_paddr)
 			sector = NULL;
 		spin_unlock(&rbio->bio_list_lock);
 		return sector;
@@ -1142,7 +1154,7 @@ static int rbio_add_io_sector(struct btrfs_raid_bio *rbio,
 			   rbio, stripe_nr);
 	ASSERT_RBIO_SECTOR(sector_nr >= 0 && sector_nr < rbio->stripe_nsectors,
 			   rbio, sector_nr);
-	ASSERT(sector->page);
+	ASSERT(sector->has_paddr);
 
 	stripe = &rbio->bioc->stripes[stripe_nr];
 	disk_start = stripe->physical + sector_nr * sectorsize;
@@ -1173,8 +1185,8 @@ static int rbio_add_io_sector(struct btrfs_raid_bio *rbio,
 		 */
 		if (last_end == disk_start && !last->bi_status &&
 		    last->bi_bdev == stripe->dev->bdev) {
-			ret = bio_add_page(last, sector->page, sectorsize,
-					   sector->pgoff);
+			ret = bio_add_page(last, phys_to_page(sector->paddr),
+					   sectorsize, offset_in_page(sector->paddr));
 			if (ret == sectorsize)
 				return 0;
 		}
@@ -1187,7 +1199,8 @@ static int rbio_add_io_sector(struct btrfs_raid_bio *rbio,
 	bio->bi_iter.bi_sector = disk_start >> SECTOR_SHIFT;
 	bio->bi_private = rbio;
 
-	__bio_add_page(bio, sector->page, sectorsize, sector->pgoff);
+	__bio_add_page(bio, phys_to_page(sector->paddr), sectorsize,
+		       offset_in_page(sector->paddr));
 	bio_list_add(bio_list, bio);
 	return 0;
 }
@@ -1205,10 +1218,8 @@ static void index_one_bio(struct btrfs_raid_bio *rbio, struct bio *bio)
 		struct sector_ptr *sector = &rbio->bio_sectors[index];
 		struct bio_vec bv = bio_iter_iovec(bio, iter);
 
-		sector->page = bv.bv_page;
-		sector->pgoff = bv.bv_offset;
-		ASSERT(sector->pgoff < PAGE_SIZE);
-
+		sector->has_paddr = true;
+		sector->paddr = bvec_phys(&bv);
 		bio_advance_iter_single(bio, &iter, sectorsize);
 		offset += sectorsize;
 	}
@@ -1288,6 +1299,15 @@ static void assert_rbio(struct btrfs_raid_bio *rbio)
 	ASSERT_RBIO(rbio->nr_data < rbio->real_stripes, rbio);
 }
 
+static inline void *kmap_local_sector(const struct sector_ptr *sector)
+{
+	/* The sector pointer must have a page mapped to it. */
+	ASSERT(sector->has_paddr);
+
+	return kmap_local_page(phys_to_page(sector->paddr)) +
+	       offset_in_page(sector->paddr);
+}
+
 /* Generate PQ for one vertical stripe. */
 static void generate_pq_vertical(struct btrfs_raid_bio *rbio, int sectornr)
 {
@@ -1300,14 +1320,13 @@ static void generate_pq_vertical(struct btrfs_raid_bio *rbio, int sectornr)
 	/* First collect one sector from each data stripe */
 	for (stripe = 0; stripe < rbio->nr_data; stripe++) {
 		sector = sector_in_rbio(rbio, stripe, sectornr, 0);
-		pointers[stripe] = kmap_local_page(sector->page) +
-				   sector->pgoff;
+		pointers[stripe] = kmap_local_sector(sector);
 	}
 
 	/* Then add the parity stripe */
 	sector = rbio_pstripe_sector(rbio, sectornr);
 	sector->uptodate = 1;
-	pointers[stripe++] = kmap_local_page(sector->page) + sector->pgoff;
+	pointers[stripe++] = kmap_local_sector(sector);
 
 	if (has_qstripe) {
 		/*
@@ -1316,8 +1335,7 @@ static void generate_pq_vertical(struct btrfs_raid_bio *rbio, int sectornr)
 		 */
 		sector = rbio_qstripe_sector(rbio, sectornr);
 		sector->uptodate = 1;
-		pointers[stripe++] = kmap_local_page(sector->page) +
-				     sector->pgoff;
+		pointers[stripe++] = kmap_local_sector(sector);
 
 		assert_rbio(rbio);
 		raid6_call.gen_syndrome(rbio->real_stripes, sectorsize,
@@ -1476,15 +1494,14 @@ static void set_rbio_range_error(struct btrfs_raid_bio *rbio, struct bio *bio)
  * stripe_pages[], thus we need to locate the sector.
  */
 static struct sector_ptr *find_stripe_sector(struct btrfs_raid_bio *rbio,
-					     struct page *page,
-					     unsigned int pgoff)
+					     phys_addr_t paddr)
 {
 	int i;
 
 	for (i = 0; i < rbio->nr_sectors; i++) {
 		struct sector_ptr *sector = &rbio->stripe_sectors[i];
 
-		if (sector->page == page && sector->pgoff == pgoff)
+		if (sector->has_paddr && sector->paddr == paddr)
 			return sector;
 	}
 	return NULL;
@@ -1504,11 +1521,10 @@ static void set_bio_pages_uptodate(struct btrfs_raid_bio *rbio, struct bio *bio)
 
 	bio_for_each_segment_all(bvec, bio, iter_all) {
 		struct sector_ptr *sector;
-		int pgoff;
+		phys_addr_t paddr = bvec_phys(bvec);
 
-		for (pgoff = bvec->bv_offset; pgoff - bvec->bv_offset < bvec->bv_len;
-		     pgoff += sectorsize) {
-			sector = find_stripe_sector(rbio, bvec->bv_page, pgoff);
+		for (u32 off = 0; off < bvec->bv_len; off += sectorsize) {
+			sector = find_stripe_sector(rbio, paddr + off);
 			ASSERT(sector);
 			if (sector)
 				sector->uptodate = 1;
@@ -1518,17 +1534,14 @@ static void set_bio_pages_uptodate(struct btrfs_raid_bio *rbio, struct bio *bio)
 
 static int get_bio_sector_nr(struct btrfs_raid_bio *rbio, struct bio *bio)
 {
-	struct bio_vec *bv = bio_first_bvec_all(bio);
+	phys_addr_t bvec_paddr = bvec_phys(bio_first_bvec_all(bio));
 	int i;
 
 	for (i = 0; i < rbio->nr_sectors; i++) {
-		struct sector_ptr *sector;
-
-		sector = &rbio->stripe_sectors[i];
-		if (sector->page == bv->bv_page && sector->pgoff == bv->bv_offset)
+		if (rbio->stripe_sectors[i].paddr == bvec_paddr)
 			break;
-		sector = &rbio->bio_sectors[i];
-		if (sector->page == bv->bv_page && sector->pgoff == bv->bv_offset)
+		if (rbio->bio_sectors[i].has_paddr &&
+		    rbio->bio_sectors[i].paddr == bvec_paddr)
 			break;
 	}
 	ASSERT(i < rbio->nr_sectors);
@@ -1810,12 +1823,10 @@ static int verify_one_sector(struct btrfs_raid_bio *rbio,
 		sector = rbio_stripe_sector(rbio, stripe_nr, sector_nr);
 	}
 
-	ASSERT(sector->page);
-
-	kaddr = kmap_local_page(sector->page) + sector->pgoff;
 	csum_expected = rbio->csum_buf +
 			(stripe_nr * rbio->stripe_nsectors + sector_nr) *
 			fs_info->csum_size;
+	kaddr = kmap_local_sector(sector);
 	ret = btrfs_check_sector_csum(fs_info, kaddr, csum_buf, csum_expected);
 	kunmap_local(kaddr);
 	return ret;
@@ -1874,9 +1885,7 @@ static int recover_vertical(struct btrfs_raid_bio *rbio, int sector_nr,
 		} else {
 			sector = rbio_stripe_sector(rbio, stripe_nr, sector_nr);
 		}
-		ASSERT(sector->page);
-		pointers[stripe_nr] = kmap_local_page(sector->page) +
-				   sector->pgoff;
+		pointers[stripe_nr] = kmap_local_sector(sector);
 		unmap_array[stripe_nr] = pointers[stripe_nr];
 	}
 
@@ -2328,7 +2337,7 @@ static bool need_read_stripe_sectors(struct btrfs_raid_bio *rbio)
 		 * thus this rbio can not be cached one, as cached one must
 		 * have all its data sectors present and uptodate.
 		 */
-		if (!sector->page || !sector->uptodate)
+		if (!sector->has_paddr || !sector->uptodate)
 			return true;
 	}
 	return false;
@@ -2518,6 +2527,7 @@ static int finish_parity_scrub(struct btrfs_raid_bio *rbio)
 	int stripe;
 	int sectornr;
 	bool has_qstripe;
+	struct page *page;
 	struct sector_ptr p_sector = { 0 };
 	struct sector_ptr q_sector = { 0 };
 	struct bio_list bio_list;
@@ -2549,29 +2559,33 @@ static int finish_parity_scrub(struct btrfs_raid_bio *rbio)
 	 */
 	clear_bit(RBIO_CACHE_READY_BIT, &rbio->flags);
 
-	p_sector.page = alloc_page(GFP_NOFS);
-	if (!p_sector.page)
+	page = alloc_page(GFP_NOFS);
+	if (!page)
 		return -ENOMEM;
-	p_sector.pgoff = 0;
+	p_sector.has_paddr = true;
+	p_sector.paddr = page_to_phys(page);
 	p_sector.uptodate = 1;
+	page = NULL;
 
 	if (has_qstripe) {
 		/* RAID6, allocate and map temp space for the Q stripe */
-		q_sector.page = alloc_page(GFP_NOFS);
-		if (!q_sector.page) {
-			__free_page(p_sector.page);
-			p_sector.page = NULL;
+		page = alloc_page(GFP_NOFS);
+		if (!page) {
+			__free_page(phys_to_page(p_sector.paddr));
+			p_sector.has_paddr = false;
 			return -ENOMEM;
 		}
-		q_sector.pgoff = 0;
+		q_sector.has_paddr = true;
+		q_sector.paddr = page_to_phys(page);
 		q_sector.uptodate = 1;
-		pointers[rbio->real_stripes - 1] = kmap_local_page(q_sector.page);
+		page = NULL;
+		pointers[rbio->real_stripes - 1] = kmap_local_sector(&q_sector);
 	}
 
 	bitmap_clear(rbio->error_bitmap, 0, rbio->nr_sectors);
 
 	/* Map the parity stripe just once */
-	pointers[nr_data] = kmap_local_page(p_sector.page);
+	pointers[nr_data] = kmap_local_sector(&p_sector);
 
 	for_each_set_bit(sectornr, &rbio->dbitmap, rbio->stripe_nsectors) {
 		struct sector_ptr *sector;
@@ -2580,8 +2594,7 @@ static int finish_parity_scrub(struct btrfs_raid_bio *rbio)
 		/* first collect one page from each data stripe */
 		for (stripe = 0; stripe < nr_data; stripe++) {
 			sector = sector_in_rbio(rbio, stripe, sectornr, 0);
-			pointers[stripe] = kmap_local_page(sector->page) +
-					   sector->pgoff;
+			pointers[stripe] = kmap_local_sector(sector);
 		}
 
 		if (has_qstripe) {
@@ -2597,7 +2610,7 @@ static int finish_parity_scrub(struct btrfs_raid_bio *rbio)
 
 		/* Check scrubbing parity and repair it */
 		sector = rbio_stripe_sector(rbio, rbio->scrubp, sectornr);
-		parity = kmap_local_page(sector->page) + sector->pgoff;
+		parity = kmap_local_sector(sector);
 		if (memcmp(parity, pointers[rbio->scrubp], sectorsize) != 0)
 			memcpy(parity, pointers[rbio->scrubp], sectorsize);
 		else
@@ -2610,12 +2623,11 @@ static int finish_parity_scrub(struct btrfs_raid_bio *rbio)
 	}
 
 	kunmap_local(pointers[nr_data]);
-	__free_page(p_sector.page);
-	p_sector.page = NULL;
-	if (q_sector.page) {
-		kunmap_local(pointers[rbio->real_stripes - 1]);
-		__free_page(q_sector.page);
-		q_sector.page = NULL;
+	__free_page(phys_to_page(p_sector.paddr));
+	p_sector.has_paddr = false;
+	if (q_sector.has_paddr) {
+		__free_page(phys_to_page(q_sector.paddr));
+		q_sector.has_paddr = false;
 	}
 
 	/*