irods: VERIFY_CHKSUM_KW failure leaves a current data object in a bad state

• master
• 4-2-stable

---

Bug Report

iRODS Version, OS and Version

4.2.7, Linux, Ubuntu bionic

What did you try to do?

rcDataObjPut a file whose MD5 is 6a57b28b63eede75341f2e06b428cf7a to a replication resource using VERIFY_CHKSUM_KW while intentionally passing an incorrect checksum of My hovercraft is full of eels.

Expected behavior

The put to fail with USER_CHKSUM_MISMATCH and the data object either not be created or be created with a stale state.

Observed behavior (including steps to reproduce, if applicable)

The put fails with USER_CHKSUM_MISMATCH and a data object is created with no checksum and size 0 bytes. The data object is marked as current (& with ils -L).

Moreover, any client subsequently running ichksum on the data object will generate a checksum of 6a57b28b63eede75341f2e06b428cf7a (MD5, or equivalent for other schemes), rather than the checksum of a file of size 0 bytes.

This leaves a data object having “no contents” while paradoxically having a correct checksum that doesn’t match the retrievable data.

Data file:

eels.txt

C program:

#include <rodsClient.h>

/*
Tested on iRODS 4.2.7

This program puts a new data object using rcDataObjPut with the
VERIFY_CHKSUM_KW parameter set to an intentionally incorrect
checksum. This simulates a checksum verification failure.

On a replication resource, one replica is left in a current (&) state
having size 0 bytes and no checksum.

I would expect there to be either no data object at all, or for the
data object to be marked stale.

Moreover, any client subsequently running ichksum on the data object
will generate a checksum of 6a57b28b63eede75341f2e06b428cf7a (MD5, or
equivalent for other schemes), rather than the checksum of a file of
size 0 bytes.

This leaves a data object having "no contents" while paradoxically
having a correct checksum that doesn't match the retrievable data.
*/

int main(int argc, char *argv[]) {
    setenv(SP_OPTION, "verify-reproc", 1);

    char* local_file = "eels.txt";

    // Real checksum is "6a57b28b63eede75341f2e06b428cf7a";
    char *local_checksum = "My hovercraft is full of eels";
    char *remote_path = "/testZone/home/irods/eels.txt";

    int status;

    rodsEnv env;
    rErrMsg_t errmsg;
    rcComm_t *conn = NULL;

    status = getRodsEnv(&env);
    if (status != 0) {
        fprintf(stderr, "Failed to load iRODS environment\n");
        return 1;
    }

    conn = rcConnect(env.rodsHost, env.rodsPort, env.rodsUserName,
                     env.rodsZone, NO_RECONN, &errmsg);
    if (!conn) {
        fprintf(stderr, "Failed to connect\n");
        return 1;
    }

    fprintf(stderr, "Connected to %s:%d zone %s as %s\n",
            env.rodsHost, env.rodsPort, env.rodsZone,
            env.rodsUserName);

    init_client_api_table();
    status = clientLogin(conn, "", "");
    if (status != 0) {
        fprintf(stderr, "Failed to login\n");
        return 1;
    }

    dataObjInp_t obj_inp;
    memset(&obj_inp, 0, sizeof obj_inp);

    obj_inp.openFlags  = O_WRONLY;
    obj_inp.createMode = 0750;
    snprintf(obj_inp.objPath, MAX_NAME_LEN, "%s", remote_path);

    addKeyVal(&obj_inp.condInput, DEF_RESC_NAME_KW, env.rodsDefResource);
    addKeyVal(&obj_inp.condInput, VERIFY_CHKSUM_KW, local_checksum);
    status = rcDataObjPut(conn, &obj_inp, local_file);
    if (status != 0) {
        char *err_subname;
        const char *err_name = rodsErrorName(status, &err_subname);
        fprintf(stderr, "Failed to put data object: error %d %s\n",
                status, err_name);

        if (status == USER_CHKSUM_MISMATCH) {
            fprintf(stderr, "Got the expected failure\n");
            status = 0;
        } else {
            fprintf(stderr, "Did not reproduce the expected failure\n");
        }
    }

    rcDisconnect(conn);

    if (status == 0) {
        return status;
    }

    return 1;
}

Makefile:

CC = gcc 
CFLAGS = -g -std=c99 -Wall -Werror 
LIBS = -lirods_client -lirods_common -lstdc++ -lssl -lcrypto -lgssapi_krb5 -lpthread -ldl -lm -lrt -ljansson

PREFIX = /opt/irods
CPPFLAGS = -I$(PREFIX)/include -I$(PREFIX)/include/irods
LDFLAGS = -L$(PREFIX)/lib -L$(PREFIX)/lib/irods/externals

all: verify-repro

verify-repro: verify-repro.c
	$(CC) $($CFLAGS) $(CPPFLAGS) -o verify-repro verify-repro.c $(LDFLAGS) $(LIBS)

test: verify-repro
	./verify-repro

clean:
	rm verify-repro

.PHONY: all test clean

Example:

$ make -f Makefile.repro test PREFIX=/home/keith/.local/miniconda/envs/baton-dev
gcc   -I/home/keith/.local/miniconda/envs/baton-dev/include -I/home/keith/.local/miniconda/envs/baton-dev/include/irods -o verify-repro verify-repro.c -L/home/keith/.local/miniconda/envs/baton-dev/lib -L/home/keith/.local/miniconda/envs/baton-dev/lib/irods/externals -lirods_client -lirods_common -lstdc++ -lssl -lcrypto -lgssapi_krb5 -lpthread -ldl -lm -lrt -ljansson

./verify-repro
Connected to localhost:1247 zone testZone as irods
Failed to put data object: error -314000 USER_CHKSUM_MISMATCH
Got the expected failure

$ ils -L
/testZone/home/irods:
  irods             0 replResc;unixfs2            0 2021-02-08.14:58 & eels.txt
        generic    /var/lib/irods/iRODS/VaultRy/home/irods/eels.txt

$ ichksum eels.txt
    eels.txt    6a57b28b63eede75341f2e06b428cf7a
Total checksum performed = 1, Failed checksum = 0

$ ils -L
/testZone/home/irods:
  irods             0 replResc;unixfs2            0 2021-02-08.14:58 & eels.txt
    6a57b28b63eede75341f2e06b428cf7a    generic    /var/lib/irods/iRODS/VaultRy/home/irods/eels.txt