errseq.c 6.6 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206
  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. /*
  7. * An errseq_t is a way of recording errors in one place, and allowing any
  8. * number of "subscribers" to tell whether it has changed since a previous
  9. * point where it was sampled.
  10. *
  11. * It's implemented as an unsigned 32-bit value. The low order bits are
  12. * designated to hold an error code (between 0 and -MAX_ERRNO). The upper bits
  13. * are used as a counter. This is done with atomics instead of locking so that
  14. * these functions can be called from any context.
  15. *
  16. * The general idea is for consumers to sample an errseq_t value. That value
  17. * can later be used to tell whether any new errors have occurred since that
  18. * sampling was done.
  19. *
  20. * Note that there is a risk of collisions if new errors are being recorded
  21. * frequently, since we have so few bits to use as a counter.
  22. *
  23. * To mitigate this, one bit is used as a flag to tell whether the value has
  24. * been sampled since a new value was recorded. That allows us to avoid bumping
  25. * the counter if no one has sampled it since the last time an error was
  26. * recorded.
  27. *
  28. * A new errseq_t should always be zeroed out. A errseq_t value of all zeroes
  29. * is the special (but common) case where there has never been an error. An all
  30. * zero value thus serves as the "epoch" if one wishes to know whether there
  31. * has ever been an error set since it was first initialized.
  32. */
  33. /* The low bits are designated for error code (max of MAX_ERRNO) */
  34. #define ERRSEQ_SHIFT ilog2(MAX_ERRNO + 1)
  35. /* This bit is used as a flag to indicate whether the value has been seen */
  36. #define ERRSEQ_SEEN (1 << ERRSEQ_SHIFT)
  37. /* The lowest bit of the counter */
  38. #define ERRSEQ_CTR_INC (1 << (ERRSEQ_SHIFT + 1))
  39. /**
  40. * errseq_set - set a errseq_t for later reporting
  41. * @eseq: errseq_t field that should be set
  42. * @err: error to set (must be between -1 and -MAX_ERRNO)
  43. *
  44. * This function sets the error in @eseq, and increments the sequence counter
  45. * if the last sequence was sampled at some point in the past.
  46. *
  47. * Any error set will always overwrite an existing error.
  48. *
  49. * Return: The previous value, primarily for debugging purposes. The
  50. * return value should not be used as a previously sampled value in later
  51. * calls as it will not have the SEEN flag set.
  52. */
  53. errseq_t errseq_set(errseq_t *eseq, int err)
  54. {
  55. errseq_t cur, old;
  56. /* MAX_ERRNO must be able to serve as a mask */
  57. BUILD_BUG_ON_NOT_POWER_OF_2(MAX_ERRNO + 1);
  58. /*
  59. * Ensure the error code actually fits where we want it to go. If it
  60. * doesn't then just throw a warning and don't record anything. We
  61. * also don't accept zero here as that would effectively clear a
  62. * previous error.
  63. */
  64. old = READ_ONCE(*eseq);
  65. if (WARN(unlikely(err == 0 || (unsigned int)-err > MAX_ERRNO),
  66. "err = %d\n", err))
  67. return old;
  68. for (;;) {
  69. errseq_t new;
  70. /* Clear out error bits and set new error */
  71. new = (old & ~(MAX_ERRNO|ERRSEQ_SEEN)) | -err;
  72. /* Only increment if someone has looked at it */
  73. if (old & ERRSEQ_SEEN)
  74. new += ERRSEQ_CTR_INC;
  75. /* If there would be no change, then call it done */
  76. if (new == old) {
  77. cur = new;
  78. break;
  79. }
  80. /* Try to swap the new value into place */
  81. cur = cmpxchg(eseq, old, new);
  82. /*
  83. * Call it success if we did the swap or someone else beat us
  84. * to it for the same value.
  85. */
  86. if (likely(cur == old || cur == new))
  87. break;
  88. /* Raced with an update, try again */
  89. old = cur;
  90. }
  91. return cur;
  92. }
  93. EXPORT_SYMBOL(errseq_set);
  94. /**
  95. * errseq_sample() - Grab current errseq_t value.
  96. * @eseq: Pointer to errseq_t to be sampled.
  97. *
  98. * This function allows callers to initialise their errseq_t variable.
  99. * If the error has been "seen", new callers will not see an old error.
  100. * If there is an unseen error in @eseq, the caller of this function will
  101. * see it the next time it checks for an error.
  102. *
  103. * Context: Any context.
  104. * Return: The current errseq value.
  105. */
  106. errseq_t errseq_sample(errseq_t *eseq)
  107. {
  108. errseq_t old = READ_ONCE(*eseq);
  109. /* If nobody has seen this error yet, then we can be the first. */
  110. if (!(old & ERRSEQ_SEEN))
  111. old = 0;
  112. return old;
  113. }
  114. EXPORT_SYMBOL(errseq_sample);
  115. /**
  116. * errseq_check() - Has an error occurred since a particular sample point?
  117. * @eseq: Pointer to errseq_t value to be checked.
  118. * @since: Previously-sampled errseq_t from which to check.
  119. *
  120. * Grab the value that eseq points to, and see if it has changed @since
  121. * the given value was sampled. The @since value is not advanced, so there
  122. * is no need to mark the value as seen.
  123. *
  124. * Return: The latest error set in the errseq_t or 0 if it hasn't changed.
  125. */
  126. int errseq_check(errseq_t *eseq, errseq_t since)
  127. {
  128. errseq_t cur = READ_ONCE(*eseq);
  129. if (likely(cur == since))
  130. return 0;
  131. return -(cur & MAX_ERRNO);
  132. }
  133. EXPORT_SYMBOL(errseq_check);
  134. /**
  135. * errseq_check_and_advance() - Check an errseq_t and advance to current value.
  136. * @eseq: Pointer to value being checked and reported.
  137. * @since: Pointer to previously-sampled errseq_t to check against and advance.
  138. *
  139. * Grab the eseq value, and see whether it matches the value that @since
  140. * points to. If it does, then just return 0.
  141. *
  142. * If it doesn't, then the value has changed. Set the "seen" flag, and try to
  143. * swap it into place as the new eseq value. Then, set that value as the new
  144. * "since" value, and return whatever the error portion is set to.
  145. *
  146. * Note that no locking is provided here for concurrent updates to the "since"
  147. * value. The caller must provide that if necessary. Because of this, callers
  148. * may want to do a lockless errseq_check before taking the lock and calling
  149. * this.
  150. *
  151. * Return: Negative errno if one has been stored, or 0 if no new error has
  152. * occurred.
  153. */
  154. int errseq_check_and_advance(errseq_t *eseq, errseq_t *since)
  155. {
  156. int err = 0;
  157. errseq_t old, new;
  158. /*
  159. * Most callers will want to use the inline wrapper to check this,
  160. * so that the common case of no error is handled without needing
  161. * to take the lock that protects the "since" value.
  162. */
  163. old = READ_ONCE(*eseq);
  164. if (old != *since) {
  165. /*
  166. * Set the flag and try to swap it into place if it has
  167. * changed.
  168. *
  169. * We don't care about the outcome of the swap here. If the
  170. * swap doesn't occur, then it has either been updated by a
  171. * writer who is altering the value in some way (updating
  172. * counter or resetting the error), or another reader who is
  173. * just setting the "seen" flag. Either outcome is OK, and we
  174. * can advance "since" and return an error based on what we
  175. * have.
  176. */
  177. new = old | ERRSEQ_SEEN;
  178. if (new != old)
  179. cmpxchg(eseq, old, new);
  180. *since = new;
  181. err = -(new & MAX_ERRNO);
  182. }
  183. return err;
  184. }
  185. EXPORT_SYMBOL(errseq_check_and_advance);