[linux-2.6-block.git] / fs / erofs / decompressor.c

// SPDX-License-Identifier: GPL-2.0-only
/*
 * Copyright (C) 2019 HUAWEI, Inc.
 *             https://www.huawei.com/
 */
#include "compress.h"
#include <linux/module.h>
#include <linux/lz4.h>

#ifndef LZ4_DISTANCE_MAX	/* history window size */
#define LZ4_DISTANCE_MAX 65535	/* set to maximum value by default */
#endif

#define LZ4_MAX_DISTANCE_PAGES	(DIV_ROUND_UP(LZ4_DISTANCE_MAX, PAGE_SIZE) + 1)
#ifndef LZ4_DECOMPRESS_INPLACE_MARGIN
#define LZ4_DECOMPRESS_INPLACE_MARGIN(srcsize)  (((srcsize) >> 8) + 32)
#endif

struct z_erofs_lz4_decompress_ctx {
	struct z_erofs_decompress_req *rq;
	/* # of encoded, decoded pages */
	unsigned int inpages, outpages;
	/* decoded block total length (used for in-place decompression) */
	unsigned int oend;
};

int z_erofs_load_lz4_config(struct super_block *sb,
			    struct erofs_super_block *dsb,
			    struct z_erofs_lz4_cfgs *lz4, int size)
{
	struct erofs_sb_info *sbi = EROFS_SB(sb);
	u16 distance;

	if (lz4) {
		if (size < sizeof(struct z_erofs_lz4_cfgs)) {
			erofs_err(sb, "invalid lz4 cfgs, size=%u", size);
			return -EINVAL;
		}
		distance = le16_to_cpu(lz4->max_distance);

		sbi->lz4.max_pclusterblks = le16_to_cpu(lz4->max_pclusterblks);
		if (!sbi->lz4.max_pclusterblks) {
			sbi->lz4.max_pclusterblks = 1;	/* reserved case */
		} else if (sbi->lz4.max_pclusterblks >
			   Z_EROFS_PCLUSTER_MAX_SIZE / EROFS_BLKSIZ) {
			erofs_err(sb, "too large lz4 pclusterblks %u",
				  sbi->lz4.max_pclusterblks);
			return -EINVAL;
		}
	} else {
		distance = le16_to_cpu(dsb->u1.lz4_max_distance);
		sbi->lz4.max_pclusterblks = 1;
	}

	sbi->lz4.max_distance_pages = distance ?
					DIV_ROUND_UP(distance, PAGE_SIZE) + 1 :
					LZ4_MAX_DISTANCE_PAGES;
	return erofs_pcpubuf_growsize(sbi->lz4.max_pclusterblks);
}

/*
 * Fill all gaps with bounce pages if it's a sparse page list. Also check if
 * all physical pages are consecutive, which can be seen for moderate CR.
 */
static int z_erofs_lz4_prepare_dstpages(struct z_erofs_lz4_decompress_ctx *ctx,
					struct page **pagepool)
{
	struct z_erofs_decompress_req *rq = ctx->rq;
	struct page *availables[LZ4_MAX_DISTANCE_PAGES] = { NULL };
	unsigned long bounced[DIV_ROUND_UP(LZ4_MAX_DISTANCE_PAGES,
					   BITS_PER_LONG)] = { 0 };
	unsigned int lz4_max_distance_pages =
				EROFS_SB(rq->sb)->lz4.max_distance_pages;
	void *kaddr = NULL;
	unsigned int i, j, top;

	top = 0;
	for (i = j = 0; i < ctx->outpages; ++i, ++j) {
		struct page *const page = rq->out[i];
		struct page *victim;

		if (j >= lz4_max_distance_pages)
			j = 0;

		/* 'valid' bounced can only be tested after a complete round */
		if (!rq->fillgaps && test_bit(j, bounced)) {
			DBG_BUGON(i < lz4_max_distance_pages);
			DBG_BUGON(top >= lz4_max_distance_pages);
			availables[top++] = rq->out[i - lz4_max_distance_pages];
		}

		if (page) {
			__clear_bit(j, bounced);
			if (!PageHighMem(page)) {
				if (!i) {
					kaddr = page_address(page);
					continue;
				}
				if (kaddr &&
				    kaddr + PAGE_SIZE == page_address(page)) {
					kaddr += PAGE_SIZE;
					continue;
				}
			}
			kaddr = NULL;
			continue;
		}
		kaddr = NULL;
		__set_bit(j, bounced);

		if (top) {
			victim = availables[--top];
			get_page(victim);
		} else {
			victim = erofs_allocpage(pagepool,
						 GFP_KERNEL | __GFP_NOFAIL);
			set_page_private(victim, Z_EROFS_SHORTLIVED_PAGE);
		}
		rq->out[i] = victim;
	}
	return kaddr ? 1 : 0;
}

static void *z_erofs_lz4_handle_overlap(struct z_erofs_lz4_decompress_ctx *ctx,
			void *inpage, unsigned int *inputmargin, int *maptype,
			bool may_inplace)
{
	struct z_erofs_decompress_req *rq = ctx->rq;
	unsigned int omargin, total, i, j;
	struct page **in;
	void *src, *tmp;

	if (rq->inplace_io) {
		omargin = PAGE_ALIGN(ctx->oend) - ctx->oend;
		if (rq->partial_decoding || !may_inplace ||
		    omargin < LZ4_DECOMPRESS_INPLACE_MARGIN(rq->inputsize))
			goto docopy;

		for (i = 0; i < ctx->inpages; ++i) {
			DBG_BUGON(rq->in[i] == NULL);
			for (j = 0; j < ctx->outpages - ctx->inpages + i; ++j)
				if (rq->out[j] == rq->in[i])
					goto docopy;
		}
	}

	if (ctx->inpages <= 1) {
		*maptype = 0;
		return inpage;
	}
	kunmap_atomic(inpage);
	might_sleep();
	src = erofs_vm_map_ram(rq->in, ctx->inpages);
	if (!src)
		return ERR_PTR(-ENOMEM);
	*maptype = 1;
	return src;

docopy:
	/* Or copy compressed data which can be overlapped to per-CPU buffer */
	in = rq->in;
	src = erofs_get_pcpubuf(ctx->inpages);
	if (!src) {
		DBG_BUGON(1);
		kunmap_atomic(inpage);
		return ERR_PTR(-EFAULT);
	}

	tmp = src;
	total = rq->inputsize;
	while (total) {
		unsigned int page_copycnt =
			min_t(unsigned int, total, PAGE_SIZE - *inputmargin);

		if (!inpage)
			inpage = kmap_atomic(*in);
		memcpy(tmp, inpage + *inputmargin, page_copycnt);
		kunmap_atomic(inpage);
		inpage = NULL;
		tmp += page_copycnt;
		total -= page_copycnt;
		++in;
		*inputmargin = 0;
	}
	*maptype = 2;
	return src;
}

/*
 * Get the exact inputsize with zero_padding feature.
 *  - For LZ4, it should work if zero_padding feature is on (5.3+);
 *  - For MicroLZMA, it'd be enabled all the time.
 */
int z_erofs_fixup_insize(struct z_erofs_decompress_req *rq, const char *padbuf,
			 unsigned int padbufsize)
{
	const char *padend;

	padend = memchr_inv(padbuf, 0, padbufsize);
	if (!padend)
		return -EFSCORRUPTED;
	rq->inputsize -= padend - padbuf;
	rq->pageofs_in += padend - padbuf;
	return 0;
}

static int z_erofs_lz4_decompress_mem(struct z_erofs_lz4_decompress_ctx *ctx,
				      u8 *out)
{
	struct z_erofs_decompress_req *rq = ctx->rq;
	bool support_0padding = false, may_inplace = false;
	unsigned int inputmargin;
	u8 *headpage, *src;
	int ret, maptype;

	DBG_BUGON(*rq->in == NULL);
	headpage = kmap_atomic(*rq->in);

	/* LZ4 decompression inplace is only safe if zero_padding is enabled */
	if (erofs_sb_has_zero_padding(EROFS_SB(rq->sb))) {
		support_0padding = true;
		ret = z_erofs_fixup_insize(rq, headpage + rq->pageofs_in,
				min_t(unsigned int, rq->inputsize,
				      EROFS_BLKSIZ - rq->pageofs_in));
		if (ret) {
			kunmap_atomic(headpage);
			return ret;
		}
		may_inplace = !((rq->pageofs_in + rq->inputsize) &
				(EROFS_BLKSIZ - 1));
	}

	inputmargin = rq->pageofs_in;
	src = z_erofs_lz4_handle_overlap(ctx, headpage, &inputmargin,
					 &maptype, may_inplace);
	if (IS_ERR(src))
		return PTR_ERR(src);

	/* legacy format could compress extra data in a pcluster. */
	if (rq->partial_decoding || !support_0padding)
		ret = LZ4_decompress_safe_partial(src + inputmargin, out,
				rq->inputsize, rq->outputsize, rq->outputsize);
	else
		ret = LZ4_decompress_safe(src + inputmargin, out,
					  rq->inputsize, rq->outputsize);

	if (ret != rq->outputsize) {
		erofs_err(rq->sb, "failed to decompress %d in[%u, %u] out[%u]",
			  ret, rq->inputsize, inputmargin, rq->outputsize);

		print_hex_dump(KERN_DEBUG, "[ in]: ", DUMP_PREFIX_OFFSET,
			       16, 1, src + inputmargin, rq->inputsize, true);
		print_hex_dump(KERN_DEBUG, "[out]: ", DUMP_PREFIX_OFFSET,
			       16, 1, out, rq->outputsize, true);

		if (ret >= 0)
			memset(out + ret, 0, rq->outputsize - ret);
		ret = -EIO;
	} else {
		ret = 0;
	}

	if (maptype == 0) {
		kunmap_atomic(headpage);
	} else if (maptype == 1) {
		vm_unmap_ram(src, ctx->inpages);
	} else if (maptype == 2) {
		erofs_put_pcpubuf(src);
	} else {
		DBG_BUGON(1);
		return -EFAULT;
	}
	return ret;
}

static int z_erofs_lz4_decompress(struct z_erofs_decompress_req *rq,
				  struct page **pagepool)
{
	struct z_erofs_lz4_decompress_ctx ctx;
	unsigned int dst_maptype;
	void *dst;
	int ret;

	ctx.rq = rq;
	ctx.oend = rq->pageofs_out + rq->outputsize;
	ctx.outpages = PAGE_ALIGN(ctx.oend) >> PAGE_SHIFT;
	ctx.inpages = PAGE_ALIGN(rq->inputsize) >> PAGE_SHIFT;

	/* one optimized fast path only for non bigpcluster cases yet */
	if (ctx.inpages == 1 && ctx.outpages == 1 && !rq->inplace_io) {
		DBG_BUGON(!*rq->out);
		dst = kmap_atomic(*rq->out);
		dst_maptype = 0;
		goto dstmap_out;
	}

	/* general decoding path which can be used for all cases */
	ret = z_erofs_lz4_prepare_dstpages(&ctx, pagepool);
	if (ret < 0) {
		return ret;
	} else if (ret > 0) {
		dst = page_address(*rq->out);
		dst_maptype = 1;
	} else {
		dst = erofs_vm_map_ram(rq->out, ctx.outpages);
		if (!dst)
			return -ENOMEM;
		dst_maptype = 2;
	}

dstmap_out:
	ret = z_erofs_lz4_decompress_mem(&ctx, dst + rq->pageofs_out);
	if (!dst_maptype)
		kunmap_atomic(dst);
	else if (dst_maptype == 2)
		vm_unmap_ram(dst, ctx.outpages);
	return ret;
}

static int z_erofs_shifted_transform(struct z_erofs_decompress_req *rq,
				     struct page **pagepool)
{
	const unsigned int nrpages_out =
		PAGE_ALIGN(rq->pageofs_out + rq->outputsize) >> PAGE_SHIFT;
	const unsigned int righthalf = min_t(unsigned int, rq->outputsize,
					     PAGE_SIZE - rq->pageofs_out);
	const unsigned int lefthalf = rq->outputsize - righthalf;
	unsigned char *src, *dst;

	if (nrpages_out > 2) {
		DBG_BUGON(1);
		return -EIO;
	}

	if (rq->out[0] == *rq->in) {
		DBG_BUGON(nrpages_out != 1);
		return 0;
	}

	src = kmap_atomic(*rq->in) + rq->pageofs_in;
	if (rq->out[0]) {
		dst = kmap_atomic(rq->out[0]);
		memcpy(dst + rq->pageofs_out, src, righthalf);
		kunmap_atomic(dst);
	}

	if (nrpages_out == 2) {
		DBG_BUGON(!rq->out[1]);
		if (rq->out[1] == *rq->in) {
			memmove(src, src + righthalf, lefthalf);
		} else {
			dst = kmap_atomic(rq->out[1]);
			memcpy(dst, src + righthalf, lefthalf);
			kunmap_atomic(dst);
		}
	}
	kunmap_atomic(src);
	return 0;
}

static struct z_erofs_decompressor decompressors[] = {
	[Z_EROFS_COMPRESSION_SHIFTED] = {
		.decompress = z_erofs_shifted_transform,
		.name = "shifted"
	},
	[Z_EROFS_COMPRESSION_LZ4] = {
		.decompress = z_erofs_lz4_decompress,
		.name = "lz4"
	},
#ifdef CONFIG_EROFS_FS_ZIP_LZMA
	[Z_EROFS_COMPRESSION_LZMA] = {
		.decompress = z_erofs_lzma_decompress,
		.name = "lzma"
	},
#endif
};

int z_erofs_decompress(struct z_erofs_decompress_req *rq,
		       struct page **pagepool)
{
	return decompressors[rq->alg].decompress(rq, pagepool);
}
Commit	Line	Data
29b24f6c	1	// SPDX-License-Identifier: GPL-2.0-only
7fc45dbc	2	/*
7fc45dbc	3	* Copyright (C) 2019 HUAWEI, Inc.
592e7cd0	4	* https://www.huawei.com/
7fc45dbc GX	5	*/
7fc45dbc GX	6	#include "compress.h"
46c2d149	7	#include <linux/module.h>
7fc45dbc GX	8	#include <linux/lz4.h>
	9
	10	#ifndef LZ4_DISTANCE_MAX /* history window size */
	11	#define LZ4_DISTANCE_MAX 65535 /* set to maximum value by default */
	12	#endif
	13
af89bcef	14	#define LZ4_MAX_DISTANCE_PAGES (DIV_ROUND_UP(LZ4_DISTANCE_MAX, PAGE_SIZE) + 1)
0ffd71bc GX	15	#ifndef LZ4_DECOMPRESS_INPLACE_MARGIN
	16	#define LZ4_DECOMPRESS_INPLACE_MARGIN(srcsize) (((srcsize) >> 8) + 32)
	17	#endif
7fc45dbc	18
d67aee76 GX	19	struct z_erofs_lz4_decompress_ctx {
	20	struct z_erofs_decompress_req *rq;
	21	/* # of encoded, decoded pages */
	22	unsigned int inpages, outpages;
	23	/* decoded block total length (used for in-place decompression) */
	24	unsigned int oend;
	25	};
	26
5d50538f	27	int z_erofs_load_lz4_config(struct super_block *sb,
46249cde GX	28	struct erofs_super_block *dsb,
46249cde GX	29	struct z_erofs_lz4_cfgs *lz4, int size)
5d50538f	30	{
4fea63f7	31	struct erofs_sb_info *sbi = EROFS_SB(sb);
46249cde GX	32	u16 distance;
	33
	34	if (lz4) {
	35	if (size < sizeof(struct z_erofs_lz4_cfgs)) {
	36	erofs_err(sb, "invalid lz4 cfgs, size=%u", size);
	37	return -EINVAL;
	38	}
	39	distance = le16_to_cpu(lz4->max_distance);
4fea63f7 GX	40
	41	sbi->lz4.max_pclusterblks = le16_to_cpu(lz4->max_pclusterblks);
	42	if (!sbi->lz4.max_pclusterblks) {
	43	sbi->lz4.max_pclusterblks = 1; /* reserved case */
	44	} else if (sbi->lz4.max_pclusterblks >
	45	Z_EROFS_PCLUSTER_MAX_SIZE / EROFS_BLKSIZ) {
	46	erofs_err(sb, "too large lz4 pclusterblks %u",
	47	sbi->lz4.max_pclusterblks);
	48	return -EINVAL;
4fea63f7	49	}
46249cde	50	} else {
14373711	51	distance = le16_to_cpu(dsb->u1.lz4_max_distance);
4fea63f7	52	sbi->lz4.max_pclusterblks = 1;
46249cde	53	}
5d50538f	54
4fea63f7	55	sbi->lz4.max_distance_pages = distance ?
5d50538f HJ	56	DIV_ROUND_UP(distance, PAGE_SIZE) + 1 :
5d50538f HJ	57	LZ4_MAX_DISTANCE_PAGES;
4fea63f7	58	return erofs_pcpubuf_growsize(sbi->lz4.max_pclusterblks);
5d50538f HJ	59	}
5d50538f HJ	60
966edfb0 GX	61	/*
	62	* Fill all gaps with bounce pages if it's a sparse page list. Also check if
	63	* all physical pages are consecutive, which can be seen for moderate CR.
	64	*/
d67aee76	65	static int z_erofs_lz4_prepare_dstpages(struct z_erofs_lz4_decompress_ctx *ctx,
eaa9172a	66	struct page **pagepool)
7fc45dbc	67	{
d67aee76	68	struct z_erofs_decompress_req *rq = ctx->rq;
7fc45dbc	69	struct page *availables[LZ4_MAX_DISTANCE_PAGES] = { NULL };
af89bcef GX	70	unsigned long bounced[DIV_ROUND_UP(LZ4_MAX_DISTANCE_PAGES,
af89bcef GX	71	BITS_PER_LONG)] = { 0 };
5d50538f HJ	72	unsigned int lz4_max_distance_pages =
5d50538f HJ	73	EROFS_SB(rq->sb)->lz4.max_distance_pages;
7fc45dbc	74	void *kaddr = NULL;
af89bcef	75	unsigned int i, j, top;
7fc45dbc	76
af89bcef	77	top = 0;
d67aee76	78	for (i = j = 0; i < ctx->outpages; ++i, ++j) {
7fc45dbc	79	struct page *const page = rq->out[i];
af89bcef	80	struct page *victim;
7fc45dbc	81
5d50538f	82	if (j >= lz4_max_distance_pages)
af89bcef GX	83	j = 0;
	84
	85	/* 'valid' bounced can only be tested after a complete round */
267f2492	86	if (!rq->fillgaps && test_bit(j, bounced)) {
5d50538f HJ	87	DBG_BUGON(i < lz4_max_distance_pages);
	88	DBG_BUGON(top >= lz4_max_distance_pages);
	89	availables[top++] = rq->out[i - lz4_max_distance_pages];
af89bcef	90	}
7fc45dbc GX	91
7fc45dbc GX	92	if (page) {
af89bcef	93	__clear_bit(j, bounced);
448b5a15 GX	94	if (!PageHighMem(page)) {
	95	if (!i) {
	96	kaddr = page_address(page);
	97	continue;
	98	}
	99	if (kaddr &&
	100	kaddr + PAGE_SIZE == page_address(page)) {
7fc45dbc	101	kaddr += PAGE_SIZE;
448b5a15 GX	102	continue;
448b5a15 GX	103	}
7fc45dbc	104	}
448b5a15	105	kaddr = NULL;
7fc45dbc GX	106	continue;
	107	}
	108	kaddr = NULL;
af89bcef	109	__set_bit(j, bounced);
7fc45dbc	110
af89bcef GX	111	if (top) {
	112	victim = availables[--top];
	113	get_page(victim);
7fc45dbc	114	} else {
b4892fa3 HJ	115	victim = erofs_allocpage(pagepool,
b4892fa3 HJ	116	GFP_KERNEL \| __GFP_NOFAIL);
6aaa7b06	117	set_page_private(victim, Z_EROFS_SHORTLIVED_PAGE);
7fc45dbc	118	}
af89bcef	119	rq->out[i] = victim;
7fc45dbc GX	120	}
	121	return kaddr ? 1 : 0;
	122	}
	123
d67aee76	124	static void z_erofs_lz4_handle_overlap(struct z_erofs_lz4_decompress_ctx ctx,
598162d0	125	void inpage, unsigned int inputmargin, int *maptype,
ab749bad	126	bool may_inplace)
7fc45dbc	127	{
d67aee76 GX	128	struct z_erofs_decompress_req *rq = ctx->rq;
d67aee76 GX	129	unsigned int omargin, total, i, j;
598162d0 GX	130	struct page **in;
	131	void src, tmp;
	132
598162d0	133	if (rq->inplace_io) {
d67aee76	134	omargin = PAGE_ALIGN(ctx->oend) - ctx->oend;
ab749bad	135	if (rq->partial_decoding \|\| !may_inplace \|\|
d67aee76	136	omargin < LZ4_DECOMPRESS_INPLACE_MARGIN(rq->inputsize))
598162d0 GX	137	goto docopy;
598162d0 GX	138
d67aee76	139	for (i = 0; i < ctx->inpages; ++i) {
598162d0	140	DBG_BUGON(rq->in[i] == NULL);
d67aee76	141	for (j = 0; j < ctx->outpages - ctx->inpages + i; ++j)
598162d0 GX	142	if (rq->out[j] == rq->in[i])
	143	goto docopy;
	144	}
	145	}
	146
d67aee76	147	if (ctx->inpages <= 1) {
598162d0 GX	148	*maptype = 0;
	149	return inpage;
	150	}
	151	kunmap_atomic(inpage);
	152	might_sleep();
d67aee76	153	src = erofs_vm_map_ram(rq->in, ctx->inpages);
598162d0 GX	154	if (!src)
	155	return ERR_PTR(-ENOMEM);
	156	*maptype = 1;
	157	return src;
	158
	159	docopy:
	160	/* Or copy compressed data which can be overlapped to per-CPU buffer */
	161	in = rq->in;
d67aee76	162	src = erofs_get_pcpubuf(ctx->inpages);
598162d0 GX	163	if (!src) {
	164	DBG_BUGON(1);
	165	kunmap_atomic(inpage);
	166	return ERR_PTR(-EFAULT);
	167	}
	168
	169	tmp = src;
	170	total = rq->inputsize;
	171	while (total) {
	172	unsigned int page_copycnt =
	173	min_t(unsigned int, total, PAGE_SIZE - *inputmargin);
	174
	175	if (!inpage)
	176	inpage = kmap_atomic(*in);
	177	memcpy(tmp, inpage + *inputmargin, page_copycnt);
	178	kunmap_atomic(inpage);
	179	inpage = NULL;
	180	tmp += page_copycnt;
	181	total -= page_copycnt;
7fc45dbc	182	++in;
598162d0	183	*inputmargin = 0;
7fc45dbc	184	}
598162d0 GX	185	*maptype = 2;
598162d0 GX	186	return src;
7fc45dbc GX	187	}
7fc45dbc GX	188
10e5f6e4 GX	189	/*
	190	* Get the exact inputsize with zero_padding feature.
	191	* - For LZ4, it should work if zero_padding feature is on (5.3+);
	192	* - For MicroLZMA, it'd be enabled all the time.
	193	*/
	194	int z_erofs_fixup_insize(struct z_erofs_decompress_req rq, const char padbuf,
	195	unsigned int padbufsize)
	196	{
	197	const char *padend;
	198
	199	padend = memchr_inv(padbuf, 0, padbufsize);
	200	if (!padend)
	201	return -EFSCORRUPTED;
	202	rq->inputsize -= padend - padbuf;
	203	rq->pageofs_in += padend - padbuf;
	204	return 0;
	205	}
	206
d67aee76	207	static int z_erofs_lz4_decompress_mem(struct z_erofs_lz4_decompress_ctx *ctx,
966edfb0	208	u8 *out)
7fc45dbc	209	{
d67aee76	210	struct z_erofs_decompress_req *rq = ctx->rq;
ab749bad	211	bool support_0padding = false, may_inplace = false;
598162d0 GX	212	unsigned int inputmargin;
598162d0 GX	213	u8 headpage, src;
598162d0	214	int ret, maptype;
7fc45dbc	215
598162d0 GX	216	DBG_BUGON(*rq->in == NULL);
598162d0 GX	217	headpage = kmap_atomic(*rq->in);
0ffd71bc	218
10e5f6e4	219	/* LZ4 decompression inplace is only safe if zero_padding is enabled */
7e508f2c	220	if (erofs_sb_has_zero_padding(EROFS_SB(rq->sb))) {
0ffd71bc	221	support_0padding = true;
10e5f6e4 GX	222	ret = z_erofs_fixup_insize(rq, headpage + rq->pageofs_in,
	223	min_t(unsigned int, rq->inputsize,
	224	EROFS_BLKSIZ - rq->pageofs_in));
	225	if (ret) {
598162d0	226	kunmap_atomic(headpage);
10e5f6e4	227	return ret;
0ffd71bc	228	}
ab749bad GX	229	may_inplace = !((rq->pageofs_in + rq->inputsize) &
ab749bad GX	230	(EROFS_BLKSIZ - 1));
0ffd71bc	231	}
7fc45dbc	232
10e5f6e4	233	inputmargin = rq->pageofs_in;
d67aee76	234	src = z_erofs_lz4_handle_overlap(ctx, headpage, &inputmargin,
ab749bad	235	&maptype, may_inplace);
598162d0 GX	236	if (IS_ERR(src))
598162d0 GX	237	return PTR_ERR(src);
7fc45dbc	238
af1038ab GX	239	/* legacy format could compress extra data in a pcluster. */
	240	if (rq->partial_decoding \|\| !support_0padding)
	241	ret = LZ4_decompress_safe_partial(src + inputmargin, out,
598162d0	242	rq->inputsize, rq->outputsize, rq->outputsize);
af1038ab GX	243	else
af1038ab GX	244	ret = LZ4_decompress_safe(src + inputmargin, out,
598162d0	245	rq->inputsize, rq->outputsize);
af1038ab	246
aa99a76b GX	247	if (ret != rq->outputsize) {
aa99a76b GX	248	erofs_err(rq->sb, "failed to decompress %d in[%u, %u] out[%u]",
598162d0	249	ret, rq->inputsize, inputmargin, rq->outputsize);
aa99a76b	250
7fc45dbc	251	print_hex_dump(KERN_DEBUG, "[ in]: ", DUMP_PREFIX_OFFSET,
598162d0	252	16, 1, src + inputmargin, rq->inputsize, true);
7fc45dbc GX	253	print_hex_dump(KERN_DEBUG, "[out]: ", DUMP_PREFIX_OFFSET,
7fc45dbc GX	254	16, 1, out, rq->outputsize, true);
aa99a76b GX	255
	256	if (ret >= 0)
	257	memset(out + ret, 0, rq->outputsize - ret);
7fc45dbc	258	ret = -EIO;
5b6e7e12 YH	259	} else {
5b6e7e12 YH	260	ret = 0;
7fc45dbc GX	261	}
7fc45dbc GX	262
598162d0	263	if (maptype == 0) {
d67aee76	264	kunmap_atomic(headpage);
598162d0	265	} else if (maptype == 1) {
d67aee76	266	vm_unmap_ram(src, ctx->inpages);
598162d0 GX	267	} else if (maptype == 2) {
	268	erofs_put_pcpubuf(src);
	269	} else {
	270	DBG_BUGON(1);
	271	return -EFAULT;
	272	}
7fc45dbc GX	273	return ret;
	274	}
	275
966edfb0	276	static int z_erofs_lz4_decompress(struct z_erofs_decompress_req *rq,
eaa9172a	277	struct page **pagepool)
7fc45dbc	278	{
d67aee76	279	struct z_erofs_lz4_decompress_ctx ctx;
7fc45dbc GX	280	unsigned int dst_maptype;
7fc45dbc GX	281	void *dst;
598162d0	282	int ret;
7fc45dbc	283
d67aee76 GX	284	ctx.rq = rq;
	285	ctx.oend = rq->pageofs_out + rq->outputsize;
	286	ctx.outpages = PAGE_ALIGN(ctx.oend) >> PAGE_SHIFT;
	287	ctx.inpages = PAGE_ALIGN(rq->inputsize) >> PAGE_SHIFT;
	288
5b6e7e12	289	/* one optimized fast path only for non bigpcluster cases yet */
d67aee76	290	if (ctx.inpages == 1 && ctx.outpages == 1 && !rq->inplace_io) {
5b6e7e12 YH	291	DBG_BUGON(!*rq->out);
	292	dst = kmap_atomic(*rq->out);
	293	dst_maptype = 0;
	294	goto dstmap_out;
7fc45dbc GX	295	}
7fc45dbc GX	296
598162d0	297	/* general decoding path which can be used for all cases */
d67aee76 GX	298	ret = z_erofs_lz4_prepare_dstpages(&ctx, pagepool);
d67aee76 GX	299	if (ret < 0) {
7fc45dbc	300	return ret;
d67aee76	301	} else if (ret > 0) {
7fc45dbc GX	302	dst = page_address(*rq->out);
7fc45dbc GX	303	dst_maptype = 1;
d67aee76 GX	304	} else {
	305	dst = erofs_vm_map_ram(rq->out, ctx.outpages);
	306	if (!dst)
	307	return -ENOMEM;
	308	dst_maptype = 2;
7fc45dbc GX	309	}
7fc45dbc GX	310
7fc45dbc	311	dstmap_out:
d67aee76	312	ret = z_erofs_lz4_decompress_mem(&ctx, dst + rq->pageofs_out);
7fc45dbc GX	313	if (!dst_maptype)
	314	kunmap_atomic(dst);
	315	else if (dst_maptype == 2)
d67aee76	316	vm_unmap_ram(dst, ctx.outpages);
7fc45dbc GX	317	return ret;
	318	}
	319
966edfb0	320	static int z_erofs_shifted_transform(struct z_erofs_decompress_req *rq,
eaa9172a	321	struct page **pagepool)
7fc45dbc GX	322	{
	323	const unsigned int nrpages_out =
	324	PAGE_ALIGN(rq->pageofs_out + rq->outputsize) >> PAGE_SHIFT;
ab749bad GX	325	const unsigned int righthalf = min_t(unsigned int, rq->outputsize,
ab749bad GX	326	PAGE_SIZE - rq->pageofs_out);
dcbe6803	327	const unsigned int lefthalf = rq->outputsize - righthalf;
7fc45dbc GX	328	unsigned char src, dst;
	329
	330	if (nrpages_out > 2) {
	331	DBG_BUGON(1);
	332	return -EIO;
	333	}
	334
	335	if (rq->out[0] == *rq->in) {
	336	DBG_BUGON(nrpages_out != 1);
	337	return 0;
	338	}
	339
ab749bad	340	src = kmap_atomic(*rq->in) + rq->pageofs_in;
4d202437	341	if (rq->out[0]) {
7fc45dbc GX	342	dst = kmap_atomic(rq->out[0]);
7fc45dbc GX	343	memcpy(dst + rq->pageofs_out, src, righthalf);
4d202437	344	kunmap_atomic(dst);
7fc45dbc GX	345	}
7fc45dbc GX	346
4d202437	347	if (nrpages_out == 2) {
7fc45dbc	348	DBG_BUGON(!rq->out[1]);
4d202437	349	if (rq->out[1] == *rq->in) {
dcbe6803	350	memmove(src, src + righthalf, lefthalf);
4d202437 GX	351	} else {
4d202437 GX	352	dst = kmap_atomic(rq->out[1]);
dcbe6803	353	memcpy(dst, src + righthalf, lefthalf);
4d202437 GX	354	kunmap_atomic(dst);
4d202437 GX	355	}
7fc45dbc	356	}
7fc45dbc GX	357	kunmap_atomic(src);
	358	return 0;
	359	}
	360
966edfb0 GX	361	static struct z_erofs_decompressor decompressors[] = {
	362	[Z_EROFS_COMPRESSION_SHIFTED] = {
	363	.decompress = z_erofs_shifted_transform,
	364	.name = "shifted"
	365	},
	366	[Z_EROFS_COMPRESSION_LZ4] = {
	367	.decompress = z_erofs_lz4_decompress,
	368	.name = "lz4"
	369	},
622ceadd GX	370	#ifdef CONFIG_EROFS_FS_ZIP_LZMA
	371	[Z_EROFS_COMPRESSION_LZMA] = {
	372	.decompress = z_erofs_lzma_decompress,
	373	.name = "lzma"
	374	},
	375	#endif
966edfb0 GX	376	};
966edfb0 GX	377
7fc45dbc	378	int z_erofs_decompress(struct z_erofs_decompress_req *rq,
eaa9172a	379	struct page **pagepool)
7fc45dbc	380	{
966edfb0	381	return decompressors[rq->alg].decompress(rq, pagepool);
7fc45dbc	382	}