[PATCH 1/2] app/testpmd: fix stack overwriting by flow command

[PATCH 1/2] app/testpmd: fix stack overwriting by flow command

[Adrien Mazarguil]

The parameter type parser function stores a stack-local address in the
global parser context when parsing a "prefix" keyword.

This usually translates to "Bad arguments" errors even for correct flow
rules as stack gets overwritten by subsequent function calls.

Fixes: d3f61b7bad20 ("app/testpmd: add flow item spec prefix length")

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
---
 app/test-pmd/cmdline_flow.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/app/test-pmd/cmdline_flow.c b/app/test-pmd/cmdline_flow.c
index 0a40005..e952c91 100644
--- a/app/test-pmd/cmdline_flow.c
+++ b/app/test-pmd/cmdline_flow.c
@@ -1808,6 +1808,8 @@ parse_vc_spec(struct context *ctx, const struct token *token,
 		return -1;
 	/* Parse parameter types. */
 	switch (ctx->curr) {
+		static const enum index prefix[] = NEXT_ENTRY(PREFIX);
+
 	case ITEM_PARAM_IS:
 		index = 0;
 		objmask = 1;
@@ -1822,7 +1824,7 @@ parse_vc_spec(struct context *ctx, const struct token *token,
 		/* Modify next token to expect a prefix. */
 		if (ctx->next_num < 2)
 			return -1;
-		ctx->next[ctx->next_num - 2] = NEXT_ENTRY(PREFIX);
+		ctx->next[ctx->next_num - 2] = prefix;
 		/* Fall through. */
 	case ITEM_PARAM_MASK:
 		index = 2;
-- 
2.1.4

[PATCH 2/2] app/testpmd: fix MAC endian issue in flow command

[Adrien Mazarguil]

MAC addresses are implicitly handled in network order since they are
actually byte strings, however this is not properly enforced with MAC masks
provided as prefix lengths, which end up inverted on little endian
systems.

Fixes: 6df81b325fa4 ("app/testpmd: add items eth/vlan to flow command")

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
---
 app/test-pmd/cmdline_flow.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/app/test-pmd/cmdline_flow.c b/app/test-pmd/cmdline_flow.c
index e952c91..0fd69f9 100644
--- a/app/test-pmd/cmdline_flow.c
+++ b/app/test-pmd/cmdline_flow.c
@@ -1057,13 +1057,13 @@ static const struct token token_list[] = {
 		.name = "dst",
 		.help = "destination MAC",
 		.next = NEXT(item_eth, NEXT_ENTRY(MAC_ADDR), item_param),
-		.args = ARGS(ARGS_ENTRY(struct rte_flow_item_eth, dst)),
+		.args = ARGS(ARGS_ENTRY_HTON(struct rte_flow_item_eth, dst)),
 	},
 	[ITEM_ETH_SRC] = {
 		.name = "src",
 		.help = "source MAC",
 		.next = NEXT(item_eth, NEXT_ENTRY(MAC_ADDR), item_param),
-		.args = ARGS(ARGS_ENTRY(struct rte_flow_item_eth, src)),
+		.args = ARGS(ARGS_ENTRY_HTON(struct rte_flow_item_eth, src)),
 	},
 	[ITEM_ETH_TYPE] = {
 		.name = "type",
@@ -2238,6 +2238,9 @@ parse_mac_addr(struct context *ctx, const struct token *token,
 	/* Bit-mask fill is not supported. */
 	if (arg->mask || size != sizeof(tmp))
 		goto error;
+	/* Only network endian is supported. */
+	if (!arg->hton)
+		goto error;
 	ret = cmdline_parse_etheraddr(NULL, str, &tmp, size);
 	if (ret < 0 || (unsigned int)ret != len)
 		goto error;
-- 
2.1.4

Re: [PATCH 2/2] app/testpmd: fix MAC endian issue in flow command

[Thomas Monjalon]

04/05/2017 19:08, Adrien Mazarguil:
> MAC addresses are implicitly handled in network order since they are
> actually byte strings, however this is not properly enforced with MAC masks
> provided as prefix lengths, which end up inverted on little endian
> systems.
> 
> Fixes: 6df81b325fa4 ("app/testpmd: add items eth/vlan to flow command")
> 
> Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

Series applies, thanks