Accept pool names and hex inode numbers in vitastor-cli rm-data

rm-data read --pool and --inode with uint64_value(), so a pool name
became 0 and the pool was silently taken from the inode's pool bits,
i.e. "--pool <other pool> --inode <full inode number>" removed data in
a different pool than requested. A hex --inode became 0 and rm-data
"removed" inode 0, reporting success.

Resolve pool names like ls does (ENOENT for an unknown pool), parse
--inode as decimal or 0x hex, and fail with EINVAL on anything else
before removing data.

docs/usage/cli.en.md

@@ -243,15 +243,16 @@ Flatten a layer, i.e. merge data and detach it from parents.
 
 ## rm-data
 
-`vitastor-cli rm-data --pool <pool> --inode <inode> [--wait-list] [--min-offset <offset>]`
+`vitastor-cli rm-data --pool <id|name> --inode <inode> [--wait-list] [--min-offset <offset>]`
 
-Remove inode data without changing metadata.
+Remove inode data without changing metadata. `<inode>` may be decimal or 0x hex.
 
 ```
 --wait-list   Retrieve full objects listings before starting to remove objects.
               Requires more memory, but allows to show correct removal progress.
 --min-offset  Purge only data starting with specified offset.
 --max-offset  Purge only data before specified offset.
+              Offsets are in bytes or with a K/M/G/T unit suffix, decimal or 0x hex.
 --client_wait_up_timeout 16  Timeout for waiting until PGs are up in seconds.
 ```
 

docs/usage/cli.ru.md

@@ -252,15 +252,16 @@ vitastor-cli dd [iimg=<image> | if=<file>] [oimg=<image> | of=<file>] [bs=1M] \
 
 ## rm-data
 
-`vitastor-cli rm-data --pool <pool> --inode <inode> [--wait-list] [--min-offset <offset>]`
+`vitastor-cli rm-data --pool <id|name> --inode <inode> [--wait-list] [--min-offset <offset>]`
 
-Удалить данные инода, не меняя метаданные образов.
+Удалить данные инода, не меняя метаданные образов. `<inode>` можно задать в десятичном виде или в hex с 0x.
 
 ```
 --wait-list   Сначала запросить полный листинг объектов, а потом начать удалять.
               Требует больше памяти, но позволяет правильно печатать прогресс удаления.
 --min-offset  Удалять только данные, начиная с заданного смещения.
 --max-offset  Удалять только данные до (исключительно) заданного смещения.
+              Смещения задаются в байтах или с суффиксом K/M/G/T, в десятичном виде или в hex с 0x.
 --client_wait_up_timeout 16  Время ожидания поднятия PG в секундах.
 ```
 

src/cmd/cli.cpp

@@ -116,12 +116,13 @@ static const char* help_text =
     "vitastor-cli flatten <layer>\n"
     "  Flatten a layer, i.e. merge data and detach it from parents.\n"
     "\n"
-    "vitastor-cli rm-data --pool <pool> --inode <inode> [--wait-list] [--min-offset <offset>]\n"
-    "  Remove inode data without changing metadata.\n"
+    "vitastor-cli rm-data --pool <id|name> --inode <inode> [--wait-list] [--min-offset <offset>]\n"
+    "  Remove inode data without changing metadata. <inode> may be decimal or 0x hex.\n"
     "  --wait-list   Retrieve full objects listings before starting to remove objects.\n"
     "                Requires more memory, but allows to show correct removal progress.\n"
     "  --min-offset  Purge only data starting with specified offset.\n"
     "  --max-offset  Purge only data before specified offset.\n"
+    "                Offsets are in bytes or with a K/M/G/T unit suffix, decimal or 0x hex.\n"
     "  --client_wait_up_timeout 16  Timeout for waiting until PGs are up in seconds.\n"
     "\n"
     "vitastor-cli merge-data <from> <to> [--target <target>]\n"

src/cmd/cli_rm_data.cpp

@@ -1,6 +1,7 @@
 // Copyright (c) Vitaliy Filippov, 2019+
 // License: VNPL-1.1 (see README.md for details)
 
+#include <ctype.h>
 #include "cli.h"
 #include "cluster_client.h"
 
@@ -23,6 +24,7 @@ struct rm_inode_t
 {
     uint64_t inode = 0;
     pool_id_t pool_id = 0;
+    std::string pool_name;
     uint64_t min_offset = 0;
     uint64_t max_offset = 0;
     bool down_ok = false;
@@ -302,6 +304,25 @@ struct rm_inode_t
             goto resume_1;
         if (state == 100)
             return;
+        if (pool_name != "")
+        {
+            pool_id = 0;
+            for (auto & pp: parent->cli->st_cli->pool_config)
+            {
+                if (pp.second.name == pool_name)
+                {
+                    pool_id = pp.first;
+                    break;
+                }
+            }
+            if (!pool_id)
+            {
+                result = (cli_result_t){ .err = ENOENT, .text = "Pool "+pool_name+" does not exist" };
+                state = 100;
+                return;
+            }
+            inode = INODE_WITH_POOL(pool_id, inode);
+        }
         if (!pool_id)
         {
             result = (cli_result_t){ .err = EINVAL, .text = "Pool is not specified" };
@@ -317,20 +338,103 @@ struct rm_inode_t
     }
 };
 
+// Decimal or 0x hex, without sign, octal or overflow
+static bool parse_rm_number(std::string s, uint64_t & value)
+{
+    int base = 10;
+    if (s.size() > 2 && s[0] == '0' && (s[1] == 'x' || s[1] == 'X'))
+    {
+        s = s.substr(2);
+        base = 16;
+    }
+    if (s == "" || s.find_first_not_of(base == 16 ? "0123456789abcdefABCDEF" : "0123456789") != std::string::npos)
+        return false;
+    errno = 0;
+    value = strtoull(s.c_str(), NULL, base);
+    return !errno;
+}
+
+// Command-line values are strings with optional K/M/G/T suffixes, internal callers pass numbers
+static bool parse_rm_offset(const json11::Json & value, uint64_t & offset)
+{
+    if (!value.is_string())
+    {
+        offset = value.uint64_value();
+        return true;
+    }
+    std::string s = value.string_value();
+    int shift = 0;
+    switch (s.size() ? tolower(s.back()) : 0)
+    {
+        case 'k': shift = 10; break;
+        case 'm': shift = 20; break;
+        case 'g': shift = 30; break;
+        case 't': shift = 40; break;
+    }
+    if (shift)
+        s.pop_back();
+    if (!parse_rm_number(s, offset) || offset > (UINT64_MAX >> shift))
+        return false;
+    offset <<= shift;
+    return true;
+}
+
+static bool parse_rm_inode(const json11::Json & value, uint64_t & inode)
+{
+    if (!value.is_string())
+    {
+        inode = value.uint64_value();
+        return true;
+    }
+    return parse_rm_number(value.string_value(), inode);
+}
+
+// A non-numeric string is a pool name, resolved after loading the pool configuration
+static bool parse_rm_pool(const json11::Json & value, uint64_t & pool_id, std::string & pool_name)
+{
+    pool_id = value.uint64_value();
+    if (value.is_string() && !pool_id && value.string_value() != "0")
+        pool_name = value.string_value();
+    return pool_id < POOL_ID_MAX && (!value.is_string() || value.string_value() != "");
+}
+
 std::function<bool(cli_result_t &)> cli_tool_t::start_rm_data(json11::Json cfg)
 {
+    uint64_t inode = 0, pool_id = 0, min_offset = 0, max_offset = 0;
+    std::string pool_name, error;
+    if (!parse_rm_inode(cfg["inode"], inode))
+        error = "Invalid --inode value: "+cfg["inode"].as_string();
+    else if (!parse_rm_pool(cfg["pool"], pool_id, pool_name))
+        error = "Invalid --pool value: "+cfg["pool"].as_string();
+    else if (!parse_rm_offset(cfg["min_offset"], min_offset))
+        error = "Invalid --min-offset value: "+cfg["min_offset"].as_string();
+    else if (!parse_rm_offset(cfg["max_offset"], max_offset))
+        error = "Invalid --max-offset value: "+cfg["max_offset"].as_string();
+    else if (!max_offset && cfg["max_offset"].is_string())
+        error = "--max-offset 0 is invalid, omit it to remove data up to the end";
+    else if (max_offset && min_offset >= max_offset)
+        error = "--min-offset must be less than --max-offset";
+    if (error != "")
+    {
+        return [error](cli_result_t & result)
+        {
+            result = (cli_result_t){ .err = EINVAL, .text = error };
+            return true;
+        };
+    }
     auto remover = new rm_inode_t();
     remover->parent = this;
-    remover->inode = cfg["inode"].uint64_value();
-    remover->pool_id = cfg["pool"].uint64_value();
+    remover->inode = inode;
+    remover->pool_id = pool_id;
+    remover->pool_name = pool_name;
     if (remover->pool_id)
     {
         remover->inode = (remover->inode & (((uint64_t)1 << (64-POOL_ID_BITS)) - 1)) | (((uint64_t)remover->pool_id) << (64-POOL_ID_BITS));
     }
     remover->down_ok = cfg["down_ok"].bool_value();
     remover->pool_id = INODE_POOL(remover->inode);
-    remover->min_offset = cfg["min_offset"].uint64_value();
-    remover->max_offset = cfg["max_offset"].uint64_value();
+    remover->min_offset = min_offset;
+    remover->max_offset = max_offset;
     return [remover](cli_result_t & result)
     {
         remover->loop();

tests/test_rm.sh

@@ -24,4 +24,132 @@ sz=$(stat -c %s ./testdata/bin/testimg.bin)
 
 [[ "$sz" = 67108864 ]] || format_error "File size is $sz bytes after resize to 64 M"
 
+# test rm-data offset parsing and validation
+
+$VITASTOR_CLI create -s 32M rmimg
+dd if=/dev/urandom of=./testdata/bin/rmimg.bin bs=1M count=32
+$VITASTOR_CLI dd iodepth=4 bs=1M if=./testdata/bin/rmimg.bin oimg=rmimg
+
+check_rmimg()
+{
+    rm -f ./testdata/bin/check.bin
+    $VITASTOR_CLI dd iodepth=4 bs=1M conv=nosparse iimg=rmimg of=./testdata/bin/check.bin
+    cmp ./testdata/bin/check.bin ./testdata/bin/rmimg.bin || format_error "$1"
+}
+
+zero_rmimg()
+{
+    dd if=/dev/zero of=./testdata/bin/rmimg.bin bs=1M seek=$1 count=$2 conv=notrunc
+}
+
+check_rmimg "Image data differs after write"
+
+RM_DATA="$VITASTOR_CLI rm-data --pool 1 --inode $($VITASTOR_CLI ls --json rmimg | jq -r '.[0].inode_num')"
+
+# Bad offsets and empty ranges must fail without removing anything
+while IFS='|' read -r -u 3 args err; do
+    if out=$($RM_DATA $args 2>&1); then
+        format_error "rm-data $args should fail"
+    fi
+    [[ "$out" = "$err" ]] || format_error "rm-data $args failed with an unexpected error: $out"
+    check_rmimg "rm-data $args removed data"
+done 3<<EOF
+--min-offset 8M --max-offset 10X|Invalid --max-offset value: 10X
+--min-offset 8M --max-offset 1.5G|Invalid --max-offset value: 1.5G
+--min-offset 8M --max-offset 0x|Invalid --max-offset value: 0x
+--min-offset 8X --max-offset 16M|Invalid --min-offset value: 8X
+--min-offset 1.5M|Invalid --min-offset value: 1.5M
+--min-offset 8M --max-offset -8M|Invalid --max-offset value: -8M
+--min-offset -1|Invalid --min-offset value: -1
+--min-offset 16777217T|Invalid --min-offset value: 16777217T
+--min-offset 8M --max-offset 16777217T|Invalid --max-offset value: 16777217T
+--min-offset 18446744073709551616|Invalid --min-offset value: 18446744073709551616
+--min-offset 8M --max-offset 0x10000000000000000|Invalid --max-offset value: 0x10000000000000000
+--min-offset 16M --max-offset 8M|--min-offset must be less than --max-offset
+--min-offset 8M --max-offset 8M|--min-offset must be less than --max-offset
+--min-offset 16M --max-offset 0|--max-offset 0 is invalid, omit it to remove data up to the end
+--max-offset 0M|--max-offset 0 is invalid, omit it to remove data up to the end
+EOF
+
+$RM_DATA --min-offset 010M --max-offset 012M
+zero_rmimg 10 2
+check_rmimg "rm-data --min-offset 010M --max-offset 012M removed wrong data"
+
+$RM_DATA --min-offset 8M --max-offset 10M
+zero_rmimg 8 2
+check_rmimg "rm-data --min-offset 8M --max-offset 10M removed wrong data"
+
+$RM_DATA --min-offset $((20*1024*1024)) --max-offset $((24*1024*1024))
+zero_rmimg 20 4
+check_rmimg "rm-data with offsets in bytes removed wrong data"
+
+$RM_DATA --min-offset 0x1a00000 --max-offset 0x1c00000
+zero_rmimg 26 2
+check_rmimg "rm-data with hex offsets removed wrong data"
+
+$RM_DATA --min-offset 30M
+zero_rmimg 30 2
+check_rmimg "rm-data --min-offset 30M removed wrong data"
+
+# test rm-data pool names and hex inode numbers
+
+$VITASTOR_CLI create-pool testpool2 -s 2 -n 4 --failure_domain osd
+wait_pool_up 60 2 2 4
+
+$VITASTOR_CLI create -s 32M -p testpool2 rmimg2
+dd if=/dev/urandom of=./testdata/bin/rmimg2.bin bs=1M count=32
+$VITASTOR_CLI dd iodepth=4 bs=1M if=./testdata/bin/rmimg2.bin oimg=rmimg2
+
+zero_rmimg2()
+{
+    dd if=/dev/zero of=./testdata/bin/rmimg2.bin bs=1M seek=$1 count=$2 conv=notrunc
+}
+
+check_rmimg2()
+{
+    check_rmimg "$1"
+    rm -f ./testdata/bin/check.bin
+    $VITASTOR_CLI dd iodepth=4 bs=1M conv=nosparse iimg=rmimg2 of=./testdata/bin/check.bin
+    cmp ./testdata/bin/check.bin ./testdata/bin/rmimg2.bin || format_error "$1"
+}
+
+check_rmimg2 "Image data differs after write"
+
+INODE1=$($VITASTOR_CLI ls --json rmimg | jq -r '.[0].inode_num')
+INODE2=$($VITASTOR_CLI ls --json rmimg2 | jq -r '.[0].inode_num')
+[[ "$($VITASTOR_CLI ls --json rmimg2 | jq -r '.[0].pool_id')" = 2 ]] || format_error "rmimg2 is not in pool 2"
+
+# Bad inode numbers and unknown pools must fail without removing anything
+while IFS='|' read -r -u 3 args err; do
+    if out=$($VITASTOR_CLI rm-data $args 2>&1); then
+        format_error "rm-data $args should fail"
+    fi
+    [[ "$out" = "$err" ]] || format_error "rm-data $args failed with an unexpected error: $out"
+    check_rmimg2 "rm-data $args removed data"
+done 3<<EOF
+--pool 2 --inode ${INODE2}x|Invalid --inode value: ${INODE2}x
+--pool 2 --inode 0x|Invalid --inode value: 0x
+--pool 2 --inode 0x0x$INODE2|Invalid --inode value: 0x0x$INODE2
+--pool 2 --inode -1|Invalid --inode value: -1
+--pool 70000 --inode $INODE2|Invalid --pool value: 70000
+--pool nosuchpool --inode $(( (1<<48) | INODE1 ))|Pool nosuchpool does not exist
+EOF
+
+$VITASTOR_CLI rm-data --pool testpool2 --inode $INODE2 --min-offset 8M --max-offset 12M
+zero_rmimg2 8 4
+check_rmimg2 "rm-data --pool testpool2 --inode $INODE2 removed wrong data"
+
+$VITASTOR_CLI rm-data --pool 2 --inode $(printf 0x%x $INODE2) --min-offset 16M --max-offset 20M
+zero_rmimg2 16 4
+check_rmimg2 "rm-data --pool 2 --inode $(printf 0x%x $INODE2) removed wrong data"
+
+$VITASTOR_CLI rm-data --inode $(printf 0x%x $(( (2<<48) | INODE2 ))) --min-offset 20M --max-offset 24M
+zero_rmimg2 20 4
+check_rmimg2 "rm-data --inode $(printf 0x%x $(( (2<<48) | INODE2 ))) removed wrong data"
+
+# --pool overrides pool bits of --inode
+$VITASTOR_CLI rm-data --pool testpool --inode $(( (2<<48) | INODE1 )) --min-offset 0 --max-offset 4M
+zero_rmimg 0 4
+check_rmimg2 "rm-data --pool testpool --inode $(( (2<<48) | INODE1 )) removed wrong data"
+
 format_green OK