1// SPDX-License-Identifier: GPL-2.0
2#include <linux/err.h>
3#include <linux/bug.h>
4#include <linux/atomic.h>
5#include <linux/errseq.h>
6#include <linux/log2.h>
7
8/*
9 * An errseq_t is a way of recording errors in one place, and allowing any
10 * number of "subscribers" to tell whether it has changed since a previous
11 * point where it was sampled.
12 *
13 * It's implemented as an unsigned 32-bit value. The low order bits are
14 * designated to hold an error code (between 0 and -MAX_ERRNO). The upper bits
15 * are used as a counter. This is done with atomics instead of locking so that
16 * these functions can be called from any context.
17 *
18 * The general idea is for consumers to sample an errseq_t value. That value
19 * can later be used to tell whether any new errors have occurred since that
20 * sampling was done.
21 *
22 * Note that there is a risk of collisions if new errors are being recorded
23 * frequently, since we have so few bits to use as a counter.
24 *
25 * To mitigate this, one bit is used as a flag to tell whether the value has
26 * been sampled since a new value was recorded. That allows us to avoid bumping
27 * the counter if no one has sampled it since the last time an error was
28 * recorded.
29 *
30 * A new errseq_t should always be zeroed out. A errseq_t value of all zeroes
31 * is the special (but common) case where there has never been an error. An all
32 * zero value thus serves as the "epoch" if one wishes to know whether there
33 * has ever been an error set since it was first initialized.
34 */
35
36/* The low bits are designated for error code (max of MAX_ERRNO) */
37#define ERRSEQ_SHIFT ilog2(MAX_ERRNO + 1)
38
39/* This bit is used as a flag to indicate whether the value has been seen */
40#define ERRSEQ_SEEN (1 << ERRSEQ_SHIFT)
41
42/* The lowest bit of the counter */
43#define ERRSEQ_CTR_INC (1 << (ERRSEQ_SHIFT + 1))
44
45/**
46 * errseq_set - set a errseq_t for later reporting
47 * @eseq: errseq_t field that should be set
48 * @err: error to set (must be between -1 and -MAX_ERRNO)
49 *
50 * This function sets the error in @eseq, and increments the sequence counter
51 * if the last sequence was sampled at some point in the past.
52 *
53 * Any error set will always overwrite an existing error.
54 *
55 * Return: The previous value, primarily for debugging purposes. The
56 * return value should not be used as a previously sampled value in later
57 * calls as it will not have the SEEN flag set.
58 */
59errseq_t errseq_set(errseq_t *eseq, int err)
60{
61 errseq_t cur, old;
62
63 /* MAX_ERRNO must be able to serve as a mask */
64 BUILD_BUG_ON_NOT_POWER_OF_2(MAX_ERRNO + 1);
65
66 /*
67 * Ensure the error code actually fits where we want it to go. If it
68 * doesn't then just throw a warning and don't record anything. We
69 * also don't accept zero here as that would effectively clear a
70 * previous error.
71 */
72 old = READ_ONCE(*eseq);
73
74 if (WARN(unlikely(err == 0 || (unsigned int)-err > MAX_ERRNO),
75 "err = %d\n", err))
76 return old;
77
78 for (;;) {
79 errseq_t new;
80
81 /* Clear out error bits and set new error */
82 new = (old & ~(MAX_ERRNO|ERRSEQ_SEEN)) | -err;
83
84 /* Only increment if someone has looked at it */
85 if (old & ERRSEQ_SEEN)
86 new += ERRSEQ_CTR_INC;
87
88 /* If there would be no change, then call it done */
89 if (new == old) {
90 cur = new;
91 break;
92 }
93
94 /* Try to swap the new value into place */
95 cur = cmpxchg(eseq, old, new);
96
97 /*
98 * Call it success if we did the swap or someone else beat us
99 * to it for the same value.
100 */
101 if (likely(cur == old || cur == new))
102 break;
103
104 /* Raced with an update, try again */
105 old = cur;
106 }
107 return cur;
108}
109EXPORT_SYMBOL(errseq_set);
110
111/**
112 * errseq_sample() - Grab current errseq_t value.
113 * @eseq: Pointer to errseq_t to be sampled.
114 *
115 * This function allows callers to initialise their errseq_t variable.
116 * If the error has been "seen", new callers will not see an old error.
117 * If there is an unseen error in @eseq, the caller of this function will
118 * see it the next time it checks for an error.
119 *
120 * Context: Any context.
121 * Return: The current errseq value.
122 */
123errseq_t errseq_sample(errseq_t *eseq)
124{
125 errseq_t old = READ_ONCE(*eseq);
126
127 /* If nobody has seen this error yet, then we can be the first. */
128 if (!(old & ERRSEQ_SEEN))
129 old = 0;
130 return old;
131}
132EXPORT_SYMBOL(errseq_sample);
133
134/**
135 * errseq_check() - Has an error occurred since a particular sample point?
136 * @eseq: Pointer to errseq_t value to be checked.
137 * @since: Previously-sampled errseq_t from which to check.
138 *
139 * Grab the value that eseq points to, and see if it has changed @since
140 * the given value was sampled. The @since value is not advanced, so there
141 * is no need to mark the value as seen.
142 *
143 * Return: The latest error set in the errseq_t or 0 if it hasn't changed.
144 */
145int errseq_check(errseq_t *eseq, errseq_t since)
146{
147 errseq_t cur = READ_ONCE(*eseq);
148
149 if (likely(cur == since))
150 return 0;
151 return -(cur & MAX_ERRNO);
152}
153EXPORT_SYMBOL(errseq_check);
154
155/**
156 * errseq_check_and_advance() - Check an errseq_t and advance to current value.
157 * @eseq: Pointer to value being checked and reported.
158 * @since: Pointer to previously-sampled errseq_t to check against and advance.
159 *
160 * Grab the eseq value, and see whether it matches the value that @since
161 * points to. If it does, then just return 0.
162 *
163 * If it doesn't, then the value has changed. Set the "seen" flag, and try to
164 * swap it into place as the new eseq value. Then, set that value as the new
165 * "since" value, and return whatever the error portion is set to.
166 *
167 * Note that no locking is provided here for concurrent updates to the "since"
168 * value. The caller must provide that if necessary. Because of this, callers
169 * may want to do a lockless errseq_check before taking the lock and calling
170 * this.
171 *
172 * Return: Negative errno if one has been stored, or 0 if no new error has
173 * occurred.
174 */
175int errseq_check_and_advance(errseq_t *eseq, errseq_t *since)
176{
177 int err = 0;
178 errseq_t old, new;
179
180 /*
181 * Most callers will want to use the inline wrapper to check this,
182 * so that the common case of no error is handled without needing
183 * to take the lock that protects the "since" value.
184 */
185 old = READ_ONCE(*eseq);
186 if (old != *since) {
187 /*
188 * Set the flag and try to swap it into place if it has
189 * changed.
190 *
191 * We don't care about the outcome of the swap here. If the
192 * swap doesn't occur, then it has either been updated by a
193 * writer who is altering the value in some way (updating
194 * counter or resetting the error), or another reader who is
195 * just setting the "seen" flag. Either outcome is OK, and we
196 * can advance "since" and return an error based on what we
197 * have.
198 */
199 new = old | ERRSEQ_SEEN;
200 if (new != old)
201 cmpxchg(eseq, old, new);
202 *since = new;
203 err = -(new & MAX_ERRNO);
204 }
205 return err;
206}
207EXPORT_SYMBOL(errseq_check_and_advance);
208

source code of linux/lib/errseq.c