On 21/8/20 19:21, Pablo Neira Ayuso wrote:
if (nla[NFTA_TABLE_USERDATA]) {
udlen = nla_len(nla[NFTA_TABLE_USERDATA]);
table->udata = kzalloc(udlen, GFP_KERNEL);
if (table->udata == NULL)
goto err_table_udata;
nla_memcpy(table->udata, nla[NFTA_TABLE_USERDATA], udlen);
table->udlen = udlen;
}
Probably this simplification instead? kzalloc() zeroes the table
object, so table->udata is NULL and ->udlen is zero.
I see. The reason why I didn't simplify at first was because when using
other nf_tables_new* functions with userdata support as reference (like
newset or newrule), these are checking for userdata length obtained via
nla_len before calling nla_memcpy.
