Update loading and chaining of the connection_limit #65
Conversation
sanfern
requested review from
jniesz,
jaysheth2,
aka320,
dalalkaran and
binojrajan
as code owners
|
This PR depends l3af-project/l3afd#279 |
sanfern
force-pushed
the
sanfern-rem-load-cl
branch
from
48d4f58 to
0d6e440
Compare
| static int is_ipv4_loopback(uint32_t *addr4); | ||
| static int is_ipv6_loopback(struct in6_addr *addr6); |
There was a problem hiding this comment.
why the inconsistent types? i.e. why not struct in_addr in line 41?
There was a problem hiding this comment.
The purpose of this method is to confirm whether the local address is a loopback address or not. It achieves this by extracting the current connection count from procfs. When it comes to IPv4, the local address is in uint32 format, whereas for IPv6 addresses, there are multiple uint32 values. Therefore, the data is organized into a struct and then passed to this verification method.
There was a problem hiding this comment.
That doesn't answer the question. Why the inconsistent types of uint32_t vs struct in6_addr? Why not simply uint8_t* in both cases? or struct in_addr and struct in6_addr?
sanfern
force-pushed
the
sanfern-rem-load-cl
branch
from
0d6e440 to
a859eb6
Compare
sanfern
force-pushed
the
sanfern-rem-load-cl
branch
2 times, most recently
from
47b3041 to
ede7129
Compare
|
|
||
| int map_fd[MAP_COUNT]; | ||
| int conn_count_map_fd = -1, tcp_conns_map_fd = -1, conn_info_map_fd = -1; |
There was a problem hiding this comment.
can we retain map_fd[] array and have enums for each index like, CONN_COUNT, TCP_CONNS & CONN_INFO so that its extensible ?
| snprintf(map_dir, MAP_PATH_SIZE, "%s/%s", map_base_dir, ifname); | ||
| struct stat st = {0}; | ||
| if (stat(map_dir, &st) != 0) { | ||
| mkdir (map_dir, 0700); |
There was a problem hiding this comment.
can we handle the error case for mkdir()
|
|
||
| memset(ports, '\0', 2048); |
There was a problem hiding this comment.
can we have MACRO for the array size ?
sanfern
force-pushed
the
sanfern-rem-load-cl
branch
2 times, most recently
from
f4f4092 to
3208c69
Compare
Signed-off-by: Santhosh Fernandes <[email protected]>
sanfern
force-pushed
the
sanfern-rem-load-cl
branch
from
3208c69 to
9fe0987
Compare
This PR is to make changes to the connection limit program to load from the l3afd. Specifically, the modification involves removing the loading and chaining functionality from the user program. The user program will no longer run as a daemon. Additionally, maps will be accessed from a directory path specific to the interface, which is located at
/sys/bpf/fs/<iface>.