xref: /linux/fs/squashfs/zstd_wrapper.c (revision 03ab8e6297acd1bc0eedaa050e2a1635c576fd11)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * Squashfs - a compressed read only filesystem for Linux
4  *
5  * Copyright (c) 2016-present, Facebook, Inc.
6  * All rights reserved.
7  *
8  * zstd_wrapper.c
9  */
10 
11 #include <linux/mutex.h>
12 #include <linux/bio.h>
13 #include <linux/slab.h>
14 #include <linux/zstd.h>
15 #include <linux/vmalloc.h>
16 
17 #include "squashfs_fs.h"
18 #include "squashfs_fs_sb.h"
19 #include "squashfs.h"
20 #include "decompressor.h"
21 #include "page_actor.h"
22 
23 struct workspace {
24 	void *mem;
25 	size_t mem_size;
26 	size_t window_size;
27 };
28 
29 static void *zstd_init(struct squashfs_sb_info *msblk, void *buff)
30 {
31 	struct workspace *wksp = kmalloc(sizeof(*wksp), GFP_KERNEL);
32 
33 	if (wksp == NULL)
34 		goto failed;
35 	wksp->window_size = max_t(size_t,
36 			msblk->block_size, SQUASHFS_METADATA_SIZE);
37 	wksp->mem_size = zstd_dstream_workspace_bound(wksp->window_size);
38 	wksp->mem = vmalloc(wksp->mem_size);
39 	if (wksp->mem == NULL)
40 		goto failed;
41 
42 	return wksp;
43 
44 failed:
45 	ERROR("Failed to allocate zstd workspace\n");
46 	kfree(wksp);
47 	return ERR_PTR(-ENOMEM);
48 }
49 
50 
51 static void zstd_free(void *strm)
52 {
53 	struct workspace *wksp = strm;
54 
55 	if (wksp)
56 		vfree(wksp->mem);
57 	kfree(wksp);
58 }
59 
60 
61 static int zstd_uncompress(struct squashfs_sb_info *msblk, void *strm,
62 	struct bio *bio, int offset, int length,
63 	struct squashfs_page_actor *output)
64 {
65 	struct workspace *wksp = strm;
66 	zstd_dstream *stream;
67 	size_t total_out = 0;
68 	int error = 0;
69 	zstd_in_buffer in_buf = { NULL, 0, 0 };
70 	zstd_out_buffer out_buf = { NULL, 0, 0 };
71 	struct bvec_iter_all iter_all = {};
72 	struct bio_vec *bvec = bvec_init_iter_all(&iter_all);
73 
74 	stream = zstd_init_dstream(wksp->window_size, wksp->mem, wksp->mem_size);
75 
76 	if (!stream) {
77 		ERROR("Failed to initialize zstd decompressor\n");
78 		return -EIO;
79 	}
80 
81 	out_buf.size = PAGE_SIZE;
82 	out_buf.dst = squashfs_first_page(output);
83 
84 	for (;;) {
85 		size_t zstd_err;
86 
87 		if (in_buf.pos == in_buf.size) {
88 			const void *data;
89 			int avail;
90 
91 			if (!bio_next_segment(bio, &iter_all)) {
92 				error = -EIO;
93 				break;
94 			}
95 
96 			avail = min(length, ((int)bvec->bv_len) - offset);
97 			data = bvec_virt(bvec);
98 			length -= avail;
99 			in_buf.src = data + offset;
100 			in_buf.size = avail;
101 			in_buf.pos = 0;
102 			offset = 0;
103 		}
104 
105 		if (out_buf.pos == out_buf.size) {
106 			out_buf.dst = squashfs_next_page(output);
107 			if (out_buf.dst == NULL) {
108 				/* Shouldn't run out of pages
109 				 * before stream is done.
110 				 */
111 				error = -EIO;
112 				break;
113 			}
114 			out_buf.pos = 0;
115 			out_buf.size = PAGE_SIZE;
116 		}
117 
118 		total_out -= out_buf.pos;
119 		zstd_err = zstd_decompress_stream(stream, &out_buf, &in_buf);
120 		total_out += out_buf.pos; /* add the additional data produced */
121 		if (zstd_err == 0)
122 			break;
123 
124 		if (zstd_is_error(zstd_err)) {
125 			ERROR("zstd decompression error: %d\n",
126 					(int)zstd_get_error_code(zstd_err));
127 			error = -EIO;
128 			break;
129 		}
130 	}
131 
132 	squashfs_finish_page(output);
133 
134 	return error ? error : total_out;
135 }
136 
137 const struct squashfs_decompressor squashfs_zstd_comp_ops = {
138 	.init = zstd_init,
139 	.free = zstd_free,
140 	.decompress = zstd_uncompress,
141 	.id = ZSTD_COMPRESSION,
142 	.name = "zstd",
143 	.supported = 1
144 };
145