diff --git a/.gitignore b/.gitignore index 3422ec7..aa0fc12 100755 --- a/.gitignore +++ b/.gitignore @@ -4,3 +4,4 @@ codex/ tests/bin +zvfs_daemon \ No newline at end of file diff --git a/README.md b/README.md index 2153729..c0a9cdb 100755 --- a/README.md +++ b/README.md @@ -179,6 +179,152 @@ LD_PRELOAD=./src/libzvfs.so ZVFS_TEST_ROOT=/zvfs ./tests/bin/hook_api_test - 相对同机 `O_DIRECT`，顺序写吞吐可有约 `2.2x~2.3x` 提升。 - 非对齐写因 RMW 开销，吞吐明显下降。 +### 5.4 fio +```shell +root@ubuntu20:/home/lian/try/zvfs# fio ./fio_script/psync.fio +test: (g=0): rw=randwrite, bs=(R) 16.0KiB-16.0KiB, (W) 16.0KiB-16.0KiB, (T) 16.0KiB-16.0KiB, ioengine=psync, iodepth=64 +fio-3.16 +Starting 1 thread +Jobs: 1 (f=1): [w(1)][100.0%][w=53.0MiB/s][w=3455 IOPS][eta 00m:00s] +test: (groupid=0, jobs=1): err= 0: pid=23035: Fri Mar 13 13:25:32 2026 + Description : ["variable bs"] + write: IOPS=3644, BW=56.9MiB/s (59.7MB/s)(570MiB/10001msec); 0 zone resets + clat (usec): min=127, max=3496, avg=272.07, stdev=91.62 + lat (usec): min=128, max=3497, avg=272.41, stdev=91.69 + clat percentiles (usec): + | 1.00th=[ 155], 5.00th=[ 169], 10.00th=[ 184], 20.00th=[ 202], + | 30.00th=[ 225], 40.00th=[ 249], 50.00th=[ 262], 60.00th=[ 277], + | 70.00th=[ 293], 80.00th=[ 322], 90.00th=[ 371], 95.00th=[ 420], + | 99.00th=[ 545], 99.50th=[ 611], 99.90th=[ 881], 99.95th=[ 1467], + | 99.99th=[ 2409] + bw ( KiB/s): min=49376, max=70387, per=99.97%, avg=58295.30, stdev=6828.37, samples=20 + iops : min= 3086, max= 4399, avg=3643.40, stdev=426.71, samples=20 + lat (usec) : 250=41.44%, 500=56.78%, 750=1.59%, 1000=0.10% + lat (msec) : 2=0.07%, 4=0.01% + cpu : usr=1.82%, sys=37.19%, ctx=54169, majf=0, minf=0 + IO depths : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0% + submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% + complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% + issued rwts: total=0,36448,0,0 short=0,0,0,0 dropped=0,0,0,0 + latency : target=0, window=0, percentile=100.00%, depth=64 + +Run status group 0 (all jobs): + WRITE: bw=56.9MiB/s (59.7MB/s), 56.9MiB/s-56.9MiB/s (59.7MB/s-59.7MB/s), io=570MiB (597MB), run=10001-10001msec + +Disk stats (read/write): + dm-0: ios=122/36085, merge=0/0, ticks=32/7748, in_queue=7780, util=99.13%, aggrios=109/36455, aggrmerge=13/12, aggrticks=28/7260, aggrin_queue=0, aggrutil=98.95% + sda: ios=109/36455, merge=13/12, ticks=28/7260, in_queue=0, util=98.95% +root@ubuntu20:/home/lian/try/zvfs# LD_PRELOAD=/home/lian/try/zvfs/src/libzvfs.so fio ./fio_script/zvfs.fio +test: (g=0): rw=randwrite, bs=(R) 16.0KiB-16.0KiB, (W) 16.0KiB-16.0KiB, (T) 16.0KiB-16.0KiB, ioengine=psync, iodepth=64 +fio-3.16 +Starting 1 thread +Jobs: 1 (f=1): [w(1)][100.0%][w=10.2MiB/s][w=650 IOPS][eta 00m:00s] +test: (groupid=0, jobs=1): err= 0: pid=23891: Fri Mar 13 13:26:54 2026 + Description : ["variable bs"] + write: IOPS=521, BW=8345KiB/s (8545kB/s)(81.5MiB/10001msec); 0 zone resets + clat (usec): min=529, max=52465, avg=1909.16, stdev=1181.14 + lat (usec): min=530, max=52467, avg=1909.72, stdev=1181.21 + clat percentiles (usec): + | 1.00th=[ 734], 5.00th=[ 922], 10.00th=[ 1037], 20.00th=[ 1237], + | 30.00th=[ 1418], 40.00th=[ 1500], 50.00th=[ 1614], 60.00th=[ 1860], + | 70.00th=[ 2024], 80.00th=[ 2311], 90.00th=[ 3130], 95.00th=[ 3982], + | 99.00th=[ 5669], 99.50th=[ 5932], 99.90th=[ 6456], 99.95th=[ 6849], + | 99.99th=[52691] + bw ( KiB/s): min= 4704, max=11200, per=99.95%, avg=8339.75, stdev=2577.19, samples=20 + iops : min= 294, max= 700, avg=521.10, stdev=161.03, samples=20 + lat (usec) : 750=1.23%, 1000=7.07% + lat (msec) : 2=60.10%, 4=26.84%, 10=4.74%, 100=0.02% + cpu : usr=0.00%, sys=3.84%, ctx=5461, majf=0, minf=7 + IO depths : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0% + submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% + complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% + issued rwts: total=0,5216,0,0 short=0,0,0,0 dropped=0,0,0,0 + latency : target=0, window=0, percentile=100.00%, depth=64 + +Run status group 0 (all jobs): + WRITE: bw=8345KiB/s (8545kB/s), 8345KiB/s-8345KiB/s (8545kB/s-8545kB/s), io=81.5MiB (85.5MB), run=10001-10001msec + +Disk stats (read/write): + dm-0: ios=39/7287, merge=0/0, ticks=12/7888, in_queue=7900, util=1.72%, aggrios=39/478, aggrmerge=0/6818, aggrticks=12/352, aggrin_queue=0, aggrutil=1.70% + sda: ios=39/478, merge=0/6818, ticks=12/352, in_queue=0, util=1.70% +``` + +#### psync +```shell +root@ubuntu20:/home/lian/try/zvfs# fio ./fio_script/psync.fio +test: (g=0): rw=randwrite, bs=(R) 16.0KiB-16.0KiB, (W) 16.0KiB-16.0KiB, (T) 16.0KiB-16.0KiB, ioengine=psync, iodepth=64 +fio-3.16 +Starting 1 thread +Jobs: 1 (f=1): [w(1)][100.0%][w=53.0MiB/s][w=3455 IOPS][eta 00m:00s] +test: (groupid=0, jobs=1): err= 0: pid=23035: Fri Mar 13 13:25:32 2026 + Description : ["variable bs"] + write: IOPS=3644, BW=56.9MiB/s (59.7MB/s)(570MiB/10001msec); 0 zone resets + clat (usec): min=127, max=3496, avg=272.07, stdev=91.62 + lat (usec): min=128, max=3497, avg=272.41, stdev=91.69 + clat percentiles (usec): + | 1.00th=[ 155], 5.00th=[ 169], 10.00th=[ 184], 20.00th=[ 202], + | 30.00th=[ 225], 40.00th=[ 249], 50.00th=[ 262], 60.00th=[ 277], + | 70.00th=[ 293], 80.00th=[ 322], 90.00th=[ 371], 95.00th=[ 420], + | 99.00th=[ 545], 99.50th=[ 611], 99.90th=[ 881], 99.95th=[ 1467], + | 99.99th=[ 2409] + bw ( KiB/s): min=49376, max=70387, per=99.97%, avg=58295.30, stdev=6828.37, samples=20 + iops : min= 3086, max= 4399, avg=3643.40, stdev=426.71, samples=20 + lat (usec) : 250=41.44%, 500=56.78%, 750=1.59%, 1000=0.10% + lat (msec) : 2=0.07%, 4=0.01% + cpu : usr=1.82%, sys=37.19%, ctx=54169, majf=0, minf=0 + IO depths : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0% + submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% + complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% + issued rwts: total=0,36448,0,0 short=0,0,0,0 dropped=0,0,0,0 + latency : target=0, window=0, percentile=100.00%, depth=64 + +Run status group 0 (all jobs): + WRITE: bw=56.9MiB/s (59.7MB/s), 56.9MiB/s-56.9MiB/s (59.7MB/s-59.7MB/s), io=570MiB (597MB), run=10001-10001msec + +Disk stats (read/write): + dm-0: ios=122/36085, merge=0/0, ticks=32/7748, in_queue=7780, util=99.13%, aggrios=109/36455, aggrmerge=13/12, aggrticks=28/7260, aggrin_queue=0, aggrutil=98.95% + sda: ios=109/36455, merge=13/12, ticks=28/7260, in_queue=0, util=98.95% +``` + +### 5.5 pgbench +```shell +root@ubuntu20:/home/lian/try/zvfs# ./scripts/run_pgbench_no_mmap.sh +当前配置： + host=127.0.0.1 port=5432 db=benchdb + scale=1 clients=1 threads=1 time=15s preload=1 + init_jobs=1 init_steps=dtg skip_init=0 + +[1/2] pgbench 初始化（-i） +some of the specified options cannot be used in initialization (-i) mode +root@ubuntu20:/home/lian/try/zvfs# ./scripts/run_pgbench_no_mmap.sh +当前配置： + host=127.0.0.1 port=5432 db=benchdb + scale=1 clients=1 threads=1 time=15s preload=1 + init_jobs=1 init_steps=dtg skip_init=0 + +[1/2] pgbench 初始化（-i） +dropping old tables... +NOTICE: table "pgbench_accounts" does not exist, skipping +NOTICE: table "pgbench_branches" does not exist, skipping +NOTICE: table "pgbench_history" does not exist, skipping +NOTICE: table "pgbench_tellers" does not exist, skipping +creating tables... +generating data... +100000 of 100000 tuples (100%) done (elapsed 0.02 s, remaining 0.00 s) +done. +[2/2] pgbench 压测（-T） +starting vacuum...end. +transaction type: +scaling factor: 1 +query mode: simple +number of clients: 1 +number of threads: 1 +duration: 15 s +number of transactions actually processed: 1381 +latency average = 10.869 ms +tps = 92.003503 (including connections establishing) +tps = 92.206743 (excluding connections establishing) +``` --- ## 6. 关键工程难点与踩坑复盘（重点） diff --git a/fio_script/kingfs b/fio_script/kingfs deleted file mode 100755 index 56af658..0000000 Binary files a/fio_script/kingfs and /dev/null differ diff --git a/fio_script/zvfs.fio b/fio_script/zvfs.fio new file mode 100644 index 0000000..3b3ccb6 --- /dev/null +++ b/fio_script/zvfs.fio @@ -0,0 +1,20 @@ + +[global] +thread=1 +group_reporting=1 +direct=1 +verify=0 +time_based=1 +runtime=10 +bs=16K +size=16384 +iodepth=64 +rw=randwrite +filename=/zvfs/fio/zvfsfio +ioengine=psync + +[test] +stonewall +description="variable bs" +bs=16K + diff --git a/kingfs b/kingfs index 120e2ea..40cac06 100644 Binary files a/kingfs and b/kingfs differ diff --git a/scripts/run_pgbench_no_mmap.sh b/scripts/run_pgbench_no_mmap.sh index c355752..2ff1da6 100755 --- a/scripts/run_pgbench_no_mmap.sh +++ b/scripts/run_pgbench_no_mmap.sh @@ -12,9 +12,13 @@ set -euo pipefail PG_HOST="127.0.0.1" PG_PORT="5432" PG_DB="benchdb" -PG_SCALE="10" -PG_TIME="20" -PG_CLIENTS="2" +PG_SCALE="1" +PG_TIME="15" +PG_CLIENTS="1" +PG_THREADS="1" +PG_INIT_JOBS="1" +PG_INIT_STEPS="dtg" +PG_SKIP_INIT="0" PG_SUPERUSER="postgres" USE_LD_PRELOAD="1" LD_PRELOAD_PATH="/home/lian/try/zvfs/src/libzvfs.so" @@ -41,13 +45,19 @@ run_pg_cmd() { echo "当前配置：" echo " host=${PG_HOST} port=${PG_PORT} db=${PG_DB}" -echo " scale=${PG_SCALE} clients=${PG_CLIENTS} time=${PG_TIME}s preload=${USE_LD_PRELOAD}" +echo " scale=${PG_SCALE} clients=${PG_CLIENTS} threads=${PG_THREADS} time=${PG_TIME}s preload=${USE_LD_PRELOAD}" +echo " init_jobs=${PG_INIT_JOBS} init_steps=${PG_INIT_STEPS} skip_init=${PG_SKIP_INIT}" echo -echo "[1/2] pgbench 初始化（-i）" -run_pg_cmd "${PG_BIN_DIR}/pgbench" \ - -h "${PG_HOST}" -p "${PG_PORT}" -i -s "${PG_SCALE}" "${PG_DB}" +if [[ "${PG_SKIP_INIT}" != "1" ]]; then + echo "[1/2] pgbench 初始化（-i）" + run_pg_cmd "${PG_BIN_DIR}/pgbench" \ + -h "${PG_HOST}" -p "${PG_PORT}" -i \ + -s "${PG_SCALE}" -I "${PG_INIT_STEPS}" "${PG_DB}" +else + echo "[1/2] 跳过初始化（PG_SKIP_INIT=1）" +fi echo "[2/2] pgbench 压测（-T）" run_pg_cmd "${PG_BIN_DIR}/pgbench" \ - -h "${PG_HOST}" -p "${PG_PORT}" -c "${PG_CLIENTS}" -T "${PG_TIME}" "${PG_DB}" + -h "${PG_HOST}" -p "${PG_PORT}" -c "${PG_CLIENTS}" -j "${PG_THREADS}" -T "${PG_TIME}" "${PG_DB}" diff --git a/src/daemon/main.c b/src/daemon/main.c index 70ae53b..15e3a3b 100644 --- a/src/daemon/main.c +++ b/src/daemon/main.c @@ -185,7 +185,7 @@ static void on_read(struct zvfs_conn *c, void *ctx) break; /* 等待更多数据 */ } - printf("[req][%s]\n", cast_opcode2string(req->opcode)); + // printf("[req][%s]\n", cast_opcode2string(req->opcode)); req->conn = c; offset += consumed; diff --git a/src/daemon/spdk_engine.c b/src/daemon/spdk_engine.c index f94c472..5e256e3 100644 --- a/src/daemon/spdk_engine.c +++ b/src/daemon/spdk_engine.c @@ -30,6 +30,17 @@ static uint64_t now_mono_ms(void) { return (uint64_t)ts.tv_sec * 1000ULL + (uint64_t)ts.tv_nsec / 1000000ULL; } +/* + * SPDK's spdk_blob_close() requires a non-NULL completion callback. + * Error paths that "fire-and-forget" close should use this no-op callback. + */ +static void blob_close_noop_cb(void *arg, int bserrno) { + (void)arg; + if (bserrno != 0) { + SPDK_ERRLOG("blob_close_noop_cb: close failed: %d\n", bserrno); + } +} + /** =========================================================== * 内部辅助：错误路径统一 push resp 并释放 req * 仅用于无法构造正常 resp 的错误情形 @@ -509,7 +520,7 @@ static void create_resize_cb(void *arg, int bserrno) { struct create_chain_ctx *cctx = arg; if (bserrno != 0) { SPDK_ERRLOG("create resize failed: %d\n", bserrno); - spdk_blob_close(cctx->blob, NULL, NULL); + spdk_blob_close(cctx->blob, blob_close_noop_cb, NULL); push_err_resp(cctx->req, bserrno); free(cctx); return; @@ -521,7 +532,7 @@ static void create_sync_cb(void *arg, int bserrno) { struct create_chain_ctx *cctx = arg; if (bserrno != 0) { SPDK_ERRLOG("create sync_md failed: %d\n", bserrno); - spdk_blob_close(cctx->blob, NULL, NULL); + spdk_blob_close(cctx->blob, blob_close_noop_cb, NULL); push_err_resp(cctx->req, bserrno); free(cctx); return; @@ -530,7 +541,7 @@ static void create_sync_cb(void *arg, int bserrno) { /* 构造 handle */ struct zvfs_blob_handle *handle = calloc(1, sizeof(*handle)); if (!handle) { - spdk_blob_close(cctx->blob, NULL, NULL); + spdk_blob_close(cctx->blob, blob_close_noop_cb, NULL); push_err_resp(cctx->req, -ENOMEM); free(cctx); return; @@ -541,7 +552,7 @@ static void create_sync_cb(void *arg, int bserrno) { atomic_init(&handle->ref_count, 1); handle->dma_buf = spdk_dma_malloc(ZVFS_DMA_BUF_SIZE, g_engine.io_unit_size, NULL); if (!handle->dma_buf) { - spdk_blob_close(cctx->blob, NULL, NULL); + spdk_blob_close(cctx->blob, blob_close_noop_cb, NULL); free(handle); push_err_resp(cctx->req, -ENOMEM); free(cctx); @@ -552,7 +563,7 @@ static void create_sync_cb(void *arg, int bserrno) { struct zvfs_resp *resp = calloc(1, sizeof(*resp)); if (!resp) { spdk_dma_free(handle->dma_buf); - spdk_blob_close(cctx->blob, NULL, NULL); + spdk_blob_close(cctx->blob, blob_close_noop_cb, NULL); free(handle); push_err_resp(cctx->req, -ENOMEM); free(cctx); @@ -566,7 +577,7 @@ static void create_sync_cb(void *arg, int bserrno) { zvfs_handle_id_t handle_id; if (engine_cache_insert(handle, &handle_id) != 0) { spdk_dma_free(handle->dma_buf); - spdk_blob_close(cctx->blob, NULL, NULL); + spdk_blob_close(cctx->blob, blob_close_noop_cb, NULL); free(handle); push_err_resp(cctx->req, -ENOMEM); free(cctx); @@ -609,7 +620,7 @@ static void blob_open_done_cb(void *arg, struct spdk_blob *blob, int bserrno) { atomic_init(&handle->ref_count, 1); handle->dma_buf = spdk_dma_malloc(ZVFS_DMA_BUF_SIZE, g_engine.io_unit_size, NULL); if (!handle->dma_buf) { - spdk_blob_close(blob, NULL, NULL); + spdk_blob_close(blob, blob_close_noop_cb, NULL); free(handle); push_err_resp(octx->req, -ENOMEM); free(octx); @@ -619,7 +630,7 @@ static void blob_open_done_cb(void *arg, struct spdk_blob *blob, int bserrno) { struct zvfs_resp *resp = calloc(1, sizeof(*resp)); if (!resp) { spdk_dma_free(handle->dma_buf); - spdk_blob_close(blob, NULL, NULL); + spdk_blob_close(blob, blob_close_noop_cb, NULL); free(handle); push_err_resp(octx->req, -ENOMEM); free(octx); @@ -632,7 +643,7 @@ static void blob_open_done_cb(void *arg, struct spdk_blob *blob, int bserrno) { zvfs_handle_id_t handle_id; if (engine_cache_insert(handle, &handle_id) != 0) { spdk_dma_free(handle->dma_buf); - spdk_blob_close(blob, NULL, NULL); + spdk_blob_close(blob, blob_close_noop_cb, NULL); free(handle); push_err_resp(octx->req, -ENOMEM); free(octx); diff --git a/src/daemon/zvfs_daemon b/src/daemon/zvfs_daemon index 37cc852..bc89231 100755 Binary files a/src/daemon/zvfs_daemon and b/src/daemon/zvfs_daemon differ diff --git a/src/hook/zvfs_hook_fcntl.c b/src/hook/zvfs_hook_fcntl.c index 186d11b..475b419 100644 --- a/src/hook/zvfs_hook_fcntl.c +++ b/src/hook/zvfs_hook_fcntl.c @@ -17,6 +17,31 @@ #include #include +static int +zvfs_fcntl_cmd_has_arg(int cmd) +{ + switch (cmd) { + case F_GETFD: + case F_GETFL: + case F_GETOWN: +#ifdef F_GETSIG + case F_GETSIG: +#endif +#ifdef F_GETLEASE + case F_GETLEASE: +#endif +#ifdef F_GETPIPE_SZ + case F_GETPIPE_SZ: +#endif +#ifdef F_GET_SEALS + case F_GET_SEALS: +#endif + return 0; + default: + return 1; + } +} + /* ------------------------------------------------------------------ */ /* 内部：fcntl 核心逻辑（已确认是 zvfs fd） */ /* ------------------------------------------------------------------ */ @@ -98,13 +123,11 @@ zvfs_fcntl_impl(int fd, int cmd, va_list ap) /* ---- 其他 cmd：透传给内核（同时维护真实 fd 状态）-------------- */ default: { - /* - * 取出可变参数作为 void* 透传。 - * 大多数 fcntl cmd 的第三个参数是 long 或指针， - * 用 void* 接收足够覆盖所有平台（64-bit）。 - */ - void *arg = va_arg(ap, void *); - return real_fcntl(fd, cmd, arg); + if (zvfs_fcntl_cmd_has_arg(cmd)) { + long arg = va_arg(ap, long); + return real_fcntl(fd, cmd, arg); + } + return real_fcntl(fd, cmd); } } /* switch */ @@ -124,13 +147,12 @@ fcntl(int fd, int cmd, ...) int ret; if (ZVFS_IN_HOOK() || !zvfs_is_zvfs_fd(fd)) { - /* - * 非 zvfs fd：透传。 - * va_list 转发需要用 vfprintf 风格，但 fcntl 没有标准的 - * va_list 版本。用 void* 提取第三参数再透传。 - */ - void *arg = va_arg(ap, void *); - ret = real_fcntl(fd, cmd, arg); + if (zvfs_fcntl_cmd_has_arg(cmd)) { + long arg = va_arg(ap, long); + ret = real_fcntl(fd, cmd, arg); + } else { + ret = real_fcntl(fd, cmd); + } va_end(ap); ZVFS_HOOK_LEAVE(); return ret; @@ -147,30 +169,29 @@ fcntl(int fd, int cmd, ...) int fcntl64(int fd, int cmd, ...) { - /* - * fcntl64 是 glibc 在 32-bit 系统上的 large-file 变体， - * 语义与 fcntl 相同，直接转发。 - */ + ZVFS_HOOK_ENTER(); va_list ap; va_start(ap, cmd); - void *arg = va_arg(ap, void *); - va_end(ap); - ZVFS_HOOK_ENTER(); int ret; if (ZVFS_IN_HOOK() || !zvfs_is_zvfs_fd(fd)) { - ret = real_fcntl64 ? real_fcntl64(fd, cmd, arg) - : real_fcntl(fd, cmd, arg); + if (zvfs_fcntl_cmd_has_arg(cmd)) { + long arg = va_arg(ap, long); + ret = real_fcntl64 ? real_fcntl64(fd, cmd, arg) + : real_fcntl(fd, cmd, arg); + } else { + ret = real_fcntl64 ? real_fcntl64(fd, cmd) + : real_fcntl(fd, cmd); + } + va_end(ap); ZVFS_HOOK_LEAVE(); return ret; } zvfs_ensure_init(); - va_list ap2; - va_start(ap2, cmd); - ret = zvfs_fcntl_impl(fd, cmd, ap2); - va_end(ap2); + ret = zvfs_fcntl_impl(fd, cmd, ap); + va_end(ap); ZVFS_HOOK_LEAVE(); return ret; diff --git a/src/hook/zvfs_hook_fd.c b/src/hook/zvfs_hook_fd.c index 937ec92..8e66b70 100644 --- a/src/hook/zvfs_hook_fd.c +++ b/src/hook/zvfs_hook_fd.c @@ -887,6 +887,10 @@ int __libc_openat64(int dirfd, const char *path, int flags, ...) int __open_2(const char *path, int flags) { + if ((flags & O_CREAT) || ((flags & O_TMPFILE) == O_TMPFILE)) { + errno = EINVAL; + return -1; + } zvfs_debug_open_log(path, NULL, "__open_2 called path=%s flags=0x%x", zvfs_dbg_str(path), flags); @@ -895,11 +899,19 @@ int __open_2(const char *path, int flags) int __open64_2(const char *path, int flags) { + if ((flags & O_CREAT) || ((flags & O_TMPFILE) == O_TMPFILE)) { + errno = EINVAL; + return -1; + } return open64(path, flags); } int __openat_2(int dirfd, const char *path, int flags) { + if ((flags & O_CREAT) || ((flags & O_TMPFILE) == O_TMPFILE)) { + errno = EINVAL; + return -1; + } zvfs_debug_open_log(path, NULL, "__openat_2 called dirfd=%d path=%s flags=0x%x", dirfd, zvfs_dbg_str(path), flags); @@ -908,6 +920,10 @@ int __openat_2(int dirfd, const char *path, int flags) int __openat64_2(int dirfd, const char *path, int flags) { + if ((flags & O_CREAT) || ((flags & O_TMPFILE) == O_TMPFILE)) { + errno = EINVAL; + return -1; + } return openat64(dirfd, path, flags); } diff --git a/src/hook/zvfs_hook_rw.c b/src/hook/zvfs_hook_rw.c index ba5be1a..fe3b341 100644 --- a/src/hook/zvfs_hook_rw.c +++ b/src/hook/zvfs_hook_rw.c @@ -382,6 +382,8 @@ zvfs_vfscanf_passthrough(FILE *stream, const char *format, va_list ap, int use_i return vfscanf(stream, format, ap); } +#if 1 +/* Disabled temporarily: fscanf hook causes db_bench crash under LD_PRELOAD. */ static int zvfs_vfscanf_impl(FILE *stream, const char *format, va_list ap, int use_isoc99) { @@ -503,6 +505,7 @@ int fscanf(FILE *stream, const char *format, ...) va_end(ap); return ret; } +#endif /* ------------------------------------------------------------------ */ /* read */ diff --git a/src/spdk_engine/io_engine.c b/src/spdk_engine/io_engine.c index 737e4f5..d3f47ad 100644 --- a/src/spdk_engine/io_engine.c +++ b/src/spdk_engine/io_engine.c @@ -16,12 +16,15 @@ struct ipc_client_ctx { int fd; - uint8_t rx_buf[ZVFS_IPC_BUF_SIZE]; + uint8_t *rx_buf; + uint8_t *tx_buf; size_t rx_len; }; static __thread struct ipc_client_ctx g_ipc_tls = { .fd = -1, + .rx_buf = NULL, + .tx_buf = NULL, .rx_len = 0, }; @@ -47,6 +50,24 @@ static void ipc_close_conn(struct ipc_client_ctx *ctx) { ctx->rx_len = 0; } +static int ipc_ensure_buffers(struct ipc_client_ctx *ctx) { + if (!ctx->rx_buf) { + ctx->rx_buf = (uint8_t *)malloc(ZVFS_IPC_BUF_SIZE); + if (!ctx->rx_buf) { + errno = ENOMEM; + return -1; + } + } + if (!ctx->tx_buf) { + ctx->tx_buf = (uint8_t *)malloc(ZVFS_IPC_BUF_SIZE); + if (!ctx->tx_buf) { + errno = ENOMEM; + return -1; + } + } + return 0; +} + static int ipc_connect(struct ipc_client_ctx *ctx) { int fd = socket(AF_UNIX, SOCK_STREAM, 0); if (fd < 0) { @@ -113,12 +134,12 @@ static int try_pop_resp(struct ipc_client_ctx *ctx, struct zvfs_resp *resp) { static int read_into_rx(struct ipc_client_ctx *ctx) { while (1) { - if (ctx->rx_len >= sizeof(ctx->rx_buf)) { + if (ctx->rx_len >= ZVFS_IPC_BUF_SIZE) { errno = EOVERFLOW; return -1; } - ssize_t n = read(ctx->fd, ctx->rx_buf + ctx->rx_len, sizeof(ctx->rx_buf) - ctx->rx_len); + ssize_t n = read(ctx->fd, ctx->rx_buf + ctx->rx_len, ZVFS_IPC_BUF_SIZE - ctx->rx_len); if (n > 0) { ctx->rx_len += (size_t)n; return 0; @@ -152,7 +173,7 @@ static int recv_one_resp(struct ipc_client_ctx *ctx, struct zvfs_resp *resp_out) return -1; } - if (ctx->rx_len == sizeof(ctx->rx_buf)) { + if (ctx->rx_len == ZVFS_IPC_BUF_SIZE) { struct zvfs_resp probe; memset(&probe, 0, sizeof(probe)); if (zvfs_deserialize_resp(ctx->rx_buf, ctx->rx_len, &probe) == 0) { @@ -183,22 +204,24 @@ static int set_errno_by_status(int status) { return -1; } -uint8_t tx[ZVFS_IPC_BUF_SIZE]; - static int ipc_do_req(struct zvfs_req *req, struct zvfs_resp *resp_out) { struct ipc_client_ctx *ctx = &g_ipc_tls; + if (ipc_ensure_buffers(ctx) != 0) { + return -1; + } + if (ipc_ensure_connected(ctx) != 0) { return -1; } - size_t tx_len = zvfs_serialize_req(req, tx, sizeof(tx)); + size_t tx_len = zvfs_serialize_req(req, ctx->tx_buf, ZVFS_IPC_BUF_SIZE); if (tx_len == 0) { errno = EMSGSIZE; return -1; } - if (write_all(ctx->fd, tx, tx_len) != 0) { + if (write_all(ctx->fd, ctx->tx_buf, tx_len) != 0) { ipc_close_conn(ctx); return -1; } diff --git a/src/zvfsmalloc.json b/src/zvfsmalloc.json index e761e99..f62036f 100755 --- a/src/zvfsmalloc.json +++ b/src/zvfsmalloc.json @@ -7,7 +7,7 @@ "method": "bdev_malloc_create", "params": { "name": "Malloc0", - "num_blocks": 1048576, + "num_blocks": 524288, "block_size": 512 } }