Skip to content

DAOS-17948 object: fix bugs for sgl duplicate logic #16795

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
gnailzenh merged 5 commits into from
Sep 24, 2025
Merged

DAOS-17948 object: fix bugs for sgl duplicate logic #16795

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
58123b4
DAOS-17948 object: fix bugs for sgl duplicate logic
wangshilong Sep 3, 2025
2a2ce08
Fix test failure
wangshilong Sep 4, 2025
8a0bd91
Merge branch 'master' of github.com:daos-stack/daos into shilongw/DAO…
wangshilong Sep 18, 2025
721c8bd
Merge branch 'master' of github.com:daos-stack/daos into shilongw/DAO…
wangshilong Sep 22, 2025
2913eed
cleanup codes
wangshilong Sep 22, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
36 changes: 18 additions & 18 deletions src/object/cli_obj.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4910,49 +4910,48 @@ obj_dup_sgls_free(struct obj_auxi_args *obj_auxi)
d_sg_list_t *sg_orig = &ctx->sgls_orig[i];
uint32_t dup_sg_idx = 0;
uint32_t dup_buf_size = 0;
uint32_t sg_nr_out = 0;
uint32_t dup_data_len = 0;
char *dup_buf;

if (!ctx->alloc_bitmaps || !ctx->alloc_bitmaps[i])
continue;

D_ASSERT(ctx->merged_bitmaps[i] != NULL);
for (j = 0; j < sg_orig->sg_nr && dup_sg_idx < sg_dup->sg_nr_out;) {
for (j = 0; j < sg_orig->sg_nr && dup_sg_idx < sg_dup->sg_nr_out; j++) {
iov = &sg_orig->sg_iovs[j];
iov_dup = &sg_dup->sg_iovs[dup_sg_idx];

if (skip_sgl_iov(false, iov)) {
j++;
if (skip_sgl_iov(false, iov))
continue;
}

/* Direct copy if entry wasn't modified */
if (!isset_range((uint8_t *)ctx->merged_bitmaps[i], j, j)) {
*iov = *iov_dup;
D_ASSERT(dup_buf_size == 0);
j++;
sg_nr_out++;
D_ASSERT(dup_data_len == 0);
dup_sg_idx++;
continue;
}
if (dup_buf_size == 0) {
if (dup_data_len == 0) {
dup_data_len = iov_dup->iov_len;
dup_buf_size = iov_dup->iov_buf_len;
dup_buf = (char *)iov_dup->iov_buf;
}

/* Update iov_len for short read */
if (dup_data_len < iov->iov_len)
iov->iov_len = dup_data_len;
/* Copy data from duplicate buffer to original buffer */
D_ASSERT(dup_buf_size >= iov->iov_buf_len);
memcpy((char *)iov->iov_buf, dup_buf, iov->iov_buf_len);
dup_buf_size -= iov->iov_buf_len;
dup_buf += iov->iov_buf_len;
sg_nr_out++;
j++;
D_ASSERT(dup_buf_size >= iov->iov_len);
memcpy((char *)iov->iov_buf, dup_buf, iov->iov_len);

@Nasf-Fan Nasf-Fan Sep 21, 2025

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Some confused here, for read case, the output lov_len depend on the read result, right? Why not copy as much data as possible as long as the output buffer is large enough? something like that:

if (iov->iov_buf_len > dup_data_len)
   iov->iov_len = dup_data_len;
else
   iov->iov_len = iov->iov_buf_len;
memcpy((char *)iov->iov_buf, dup_buf, iov->iov_len);

On the other hand, it the application requires only copy specified length to the given buffer which real size is large than the required length (iov->iov_len < iov->iov_buf_len), then how can we guarantee that all the fetched data can be filled into the given SGL buffer? I may miss something.

@wangshilong wangshilong Sep 22, 2025

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is what I currently follow: I think daos will not return more data than requested iov_len, iov_buf_len shall be equal or larger than iov_len, and when copying data back, real fetched data might be less than user requested iov_len. and it will not return more data then requested iov_len.

dup_data_len -= iov->iov_len;
dup_buf += iov->iov_len;
dup_buf_size -= iov->iov_len;

/* When current duplicate buffer is exhausted, get next entry */
if (dup_buf_size == 0)
if (dup_data_len == 0)
dup_sg_idx++;
}
sg_orig->sg_nr_out = sg_nr_out;
sg_orig->sg_nr_out = j;
}
}

Expand All @@ -4977,7 +4976,8 @@ obj_reasb_io_fini(struct obj_auxi_args *obj_auxi, bool retry)
}
obj_bulk_fini(obj_auxi);
obj_auxi_free_failed_tgt_list(obj_auxi);
obj_dup_sgls_free(obj_auxi);
if (!retry)
obj_dup_sgls_free(obj_auxi);
obj_reasb_req_fini(&obj_auxi->reasb_req, obj_auxi->iod_nr);
obj_auxi->req_reasbed = false;

Expand Down
78 changes: 78 additions & 0 deletions src/tests/suite/daos_obj_array.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1729,6 +1729,83 @@ array_recx_write_mixed_iov_read_single(void **state)
daos_obj_close(oh, NULL);
}

#define ONE_KB (1024)
#define TWO_MB (2 * 1024 * 1024)

void
test_sgl_merge_short_read(void **state)
{
test_arg_t *arg = *state;
daos_obj_id_t oid;
daos_handle_t oh;
d_iov_t dkey;
d_sg_list_t sgl;
daos_iod_t iod;
daos_recx_t recx;
char *wbuf, *rbuf;
d_iov_t *sg_iovs = NULL;
int total_iovs = TWO_MB / ONE_KB;
int data_offset = 0;
int rc, i;

oid = daos_test_oid_gen(arg->coh, OC_SX, 0, 0, arg->myrank);
rc = daos_obj_open(arg->coh, oid, DAOS_OO_RW, &oh, NULL);
assert_rc_equal(rc, 0);

/* Buffer allocation with guard pages */
wbuf = malloc(ONE_MB);
assert_true(wbuf != NULL);
dts_buf_render(wbuf, ONE_MB);

/* Allocate 2MB read buffer with guard pages */
rbuf = calloc(1, TWO_MB + FOUR_KB);
assert_true(rbuf != NULL);

/* Build scatter-gather list for 1KB chunks */
sg_iovs = calloc(total_iovs, sizeof(d_iov_t));
assert_true(sg_iovs != NULL);

/* 1MB data read in 1KB chunks */
for (i = 0; i < total_iovs; i++) {
d_iov_set(&sg_iovs[data_offset], rbuf + i * ONE_KB, ONE_KB);
data_offset++;
}

/* Configure scatter-gather list */
sgl = (d_sg_list_t){.sg_nr = total_iovs, .sg_nr_out = 0, .sg_iovs = sg_iovs};

/* Prepare I/O descriptor */
d_iov_set(&dkey, "short_key", 9);
d_iov_set(&iod.iod_name, "big_akey", 8);
iod.iod_nr = 1;
iod.iod_size = 1;
iod.iod_type = DAOS_IOD_ARRAY;
iod.iod_recxs = &recx;
recx.rx_idx = 0;
recx.rx_nr = ONE_MB - 4;

/* Write initial data with single IOV */
d_sg_list_t write_sgl = {
.sg_nr = 1,
.sg_iovs = &(d_iov_t){.iov_buf = wbuf, .iov_len = ONE_MB - 4, .iov_buf_len = ONE_MB}};
rc = daos_obj_update(oh, DAOS_TX_NONE, 0, &dkey, 1, &iod, &write_sgl, NULL);
assert_rc_equal(rc, 0);

/* Read back using 1KB chunks */
rc = daos_obj_fetch(oh, DAOS_TX_NONE, 0, &dkey, 1, &iod, &sgl, NULL, NULL);
assert_rc_equal(rc, 0);
assert_rc_equal(sgl.sg_nr_out, 1024);
assert_rc_equal(sg_iovs[1023].iov_len, 1020);

/* Verify only 1MB - 4 was read successfully */
assert_memory_equal(wbuf, rbuf, ONE_MB - 4);

free(wbuf);
free(rbuf);
free(sg_iovs);
daos_obj_close(oh, NULL);
}

static const struct CMUnitTest array_tests[] = {
{"ARRAY0: small_sgl", small_sgl, NULL, test_case_teardown},
{"ARRAY1: byte array with buffer on stack", byte_array_simple_stack, NULL, test_case_teardown},
Expand Down Expand Up @@ -1759,6 +1836,7 @@ static const struct CMUnitTest array_tests[] = {
{"ARRAY20: recx read mixed iov", array_recx_read_mixed_iov, NULL, test_case_teardown},
{"ARRAY21: recx write mixed iov", array_recx_write_mixed_iov_read_single, NULL,
test_case_teardown},
{"ARRAY22: recx short read", test_sgl_merge_short_read, NULL, test_case_teardown},
};

static int
Expand Down
Loading