2022-04-04 Korin777

測驗 1

void *memchr_opt(const void *src_void, int c, size_t length)
{
    const unsigned char *src = (const unsigned char *) src_void;
    unsigned char d = c;

    while (UNALIGNED(src)) {
        if (!length--)
            return NULL;
        if (*src == d)
            return (void *) src;
        src++;
    }

    if (!TOO_SMALL(length)) {
        /* If we get this far, we know that length is large and
         * src is word-aligned.
         */

        /* The fast code reads the source one word at a time and only performs
         * the bytewise search on word-sized segments if they contain the search
         * character, which is detected by XORing the word-sized segment with a
         * word-sized block of the search character and then detecting for the
         * presence of NULL in the result.
         */
        unsigned long *asrc = (unsigned long *) src;
        unsigned long mask = d << 8 | d;
        mask = mask << 16 | mask;
        for (unsigned int i = 32; i < LBLOCKSIZE * 8; i <<= 1)
            mask = (mask << i) | mask;

        while (length >= LBLOCKSIZE) {
            // DETECT_CHAR macro return nonezero if asrc caontain d  
            unsigned long check = DETECT_CHAR(*asrc,mask); 
            if(check) {
                // return the address of the first encouter d 
                return (void *) asrc + (ffsl(check) >> 3) - 1;
            }
            // update asrc and length
            asrc = asrc + 1;
            length -= LBLOCKSIZE;
        }

        /* If there are fewer than LBLOCKSIZE characters left, then we resort to
         * the bytewise loop.
         */
        src = (unsigned char *) asrc;
    }

    while (length--) {
        if (*src == d)
            return (void *) src;
        src++;
    }

    return NULL;
}

測驗 2

static inline ringidx_t cond_reload(ringidx_t idx, const ringidx_t *loc)
{
    ringidx_t fresh = __atomic_load_n(loc, __ATOMIC_RELAXED);
    if (before(idx, fresh)) { /* fresh is after idx, use this instead */
        idx = fresh;
    } else { /* Continue with next slot */
        // since old.e.idx != tail - size ,  current slot is used , so use next slot
        idx++; // DDD
    }
    return idx;
}
static inline ringidx_t find_tail(lfring_t *lfr, ringidx_t head, ringidx_t tail)
{
    if (lfr->flags & LFRING_FLAG_SP) /* single-producer enqueue */
        return __atomic_load_n(&lfr->tail, __ATOMIC_ACQUIRE);

    /* Multi-producer enqueue.
     * Scan ring for new elements that have been written but not released.
     */
    ringidx_t mask = lfr->mask;
    // mask = size - 1 , so size = mask + 1
    ringidx_t size = mask + 1; // KKK
    while (before(tail, head + size) &&
           // lfr->ring[tail & mask].idx == tail means this slot is used , so use next slot
            __atomic_load_n(&lfr->ring[tail & mask].idx, __ATOMIC_ACQUIRE) == // TTT
               tail)
        tail++;
    tail = cond_update(&lfr->tail, tail);
    return tail;
}
uint32_t lfring_dequeue(lfring_t *lfr,
                        void **restrict elems,
                        uint32_t n_elems,
                        uint32_t *index)
{
    ringidx_t mask = lfr->mask;
    intptr_t actual;
    ringidx_t head = __atomic_load_n(&lfr->head, __ATOMIC_RELAXED);
    ringidx_t tail = __atomic_load_n(&lfr->tail, __ATOMIC_ACQUIRE);
    do {
        actual = MIN((intptr_t)(tail - head), (intptr_t) n_elems);
        if (UNLIKELY(actual <= 0)) {
            /* Ring buffer is empty, scan for new but unreleased elements */
            tail = find_tail(lfr, head, tail);
            actual = MIN((intptr_t)(tail - head), (intptr_t) n_elems);
            if (actual <= 0)
                return 0;
        }
        for (uint32_t i = 0; i < (uint32_t) actual; i++)
            elems[i] = lfr->ring[(head + i) & mask].ptr;
        smp_fence(LoadStore);                        // Order loads only
        if (UNLIKELY(lfr->flags & LFRING_FLAG_SC)) { /* Single-consumer */
            __atomic_store_n(&lfr->head, head + actual, __ATOMIC_RELAXED);
            break;
        }

        /* else: lock-free multi-consumer */
    } while (!__atomic_compare_exchange_n(
        &lfr->head, &head, /* Updated on failure */
        __atomic_load_n(&lfr->head, __ATOMIC_ACQUIRE) + actual, // HHH
        /* weak */ false, __ATOMIC_RELAXED, __ATOMIC_RELAXED));
    // if head == lfr->head , lfr->head isn't modified by other consumer , write lfr->head + actual into lfr->head and leave loop
    // if head != lfr->head , lfr->head is modified by other consumer , write lfr->head into head and continue loop

    *index = (uint32_t) head;
    return (uint32_t) actual;
}

測驗 3

static void periodic_routine(struct work_struct *ws)
{
    // if loded kill tracer and tracee
    if (likely(loaded))
        check();
    // queue new work on wq after JIFFIES_DELAY
	queue_delayed_work(wq, &dont_trace_task, JIFFIES_DELAY);
}