)]}'
{"src/openvpn/mudp.c":[{"author":{"_account_id":1000053,"name":"Răzvan Cojocaru","display_name":"Razvan Cojocaru","email":"razvanc@mailbox.org","username":"razvanc"},"change_message_id":"5c45c285f6b58c5c6d7b25d76ad1bbb103002a0f","unresolved":true,"context_lines":[{"line_number":486,"context_line":"    /* UDP port ready to accept write */"},{"line_number":487,"context_line":"    if (rwflags \u0026 SOCKET_WRITE)"},{"line_number":488,"context_line":"    {"},{"line_number":489,"context_line":"        multi_process_outgoing_link(m, mpp_flags);"},{"line_number":490,"context_line":"    }"},{"line_number":491,"context_line":"    /* Incoming data on UDP port */"},{"line_number":492,"context_line":"    else if (rwflags \u0026 SOCKET_READ)"}],"source_content_type":"text/x-csrc","patch_set":1,"id":"2ca8d0ff_900a37cc","side":"PARENT","line":489,"updated":"2026-10-27 11:36:59.000000000","message":"This used to do nothing if `m-\u003epending` was `NULL`.\n\nNow we are going through: `multi_io_write_link()` -\u003e `multi_io_write_link_pre()` where if `m-\u003epending` is `NULL` we just take the next packet from `m-\u003embuf`, whatever that is (TCP as well?).\n\nIt\u0027s not fully clear to me how likely that is to be a problem, but it also doesn\u0027t seem impossible to occur.","commit_id":"0ad283c0198adbac3713822bd4c00ef5a615d7b6"},{"author":{"_account_id":1000034,"name":"Gianmarco De Gregori","display_name":"Gianmarco De Gregori","email":"gianmarco@mandelbit.com","username":"its_Giaan"},"change_message_id":"f1abe83e14754e342d907edf9ad97584c50e31b7","unresolved":false,"context_lines":[{"line_number":486,"context_line":"    /* UDP port ready to accept write */"},{"line_number":487,"context_line":"    if (rwflags \u0026 SOCKET_WRITE)"},{"line_number":488,"context_line":"    {"},{"line_number":489,"context_line":"        multi_process_outgoing_link(m, mpp_flags);"},{"line_number":490,"context_line":"    }"},{"line_number":491,"context_line":"    /* Incoming data on UDP port */"},{"line_number":492,"context_line":"    else if (rwflags \u0026 SOCKET_READ)"}],"source_content_type":"text/x-csrc","patch_set":1,"id":"a54930dd_52a113e8","side":"PARENT","line":489,"in_reply_to":"2ca8d0ff_900a37cc","updated":"2026-11-03 08:58:13.000000000","message":"Thanks for spotting this. I added a part to this patch to handle exactly this\ncase, TA_SOCKET_WRITE no longer falls back to the bcast/c2c queue, and\nthe drain extracts the queued packets itself with multi_get_queue().","commit_id":"0ad283c0198adbac3713822bd4c00ef5a615d7b6"}],"src/openvpn/multi_io.c":[{"author":{"_account_id":1000053,"name":"Răzvan Cojocaru","display_name":"Razvan Cojocaru","email":"razvanc@mailbox.org","username":"razvanc"},"change_message_id":"ae978127b80fd5f41c6566ca1bc9a704fe644802","unresolved":true,"context_lines":[{"line_number":686,"context_line":"     */"},{"line_number":687,"context_line":"    {"},{"line_number":688,"context_line":"        struct multi_instance *mi;"},{"line_number":689,"context_line":"        while (!IS_SIG(\u0026m-\u003etop) \u0026\u0026 (mi \u003d multi_get_queue(m-\u003embuf)) !\u003d NULL)"},{"line_number":690,"context_line":"        {"},{"line_number":691,"context_line":"            multi_io_action(m, mi, NULL, TA_SOCKET_WRITE, true);"},{"line_number":692,"context_line":"        }"}],"source_content_type":"text/x-csrc","patch_set":2,"id":"aa7e4410_d79a5417","line":689,"updated":"2026-11-03 09:52:14.000000000","message":"Before the patch, `multi_get_queue()` ran in `multi_io_write_link_pre()` only when `m-\u003epending` was `NULL`, but now it runs every time.\n\nCould that end up losing packets?","commit_id":"be5e142bdaa1060f0fe42dc8f882a913d8a4e378"},{"author":{"_account_id":1000053,"name":"Răzvan Cojocaru","display_name":"Razvan Cojocaru","email":"razvanc@mailbox.org","username":"razvanc"},"change_message_id":"11037f3b60a4218db79806592e3d46359eee962b","unresolved":false,"context_lines":[{"line_number":686,"context_line":"     */"},{"line_number":687,"context_line":"    {"},{"line_number":688,"context_line":"        struct multi_instance *mi;"},{"line_number":689,"context_line":"        while (!IS_SIG(\u0026m-\u003etop) \u0026\u0026 (mi \u003d multi_get_queue(m-\u003embuf)) !\u003d NULL)"},{"line_number":690,"context_line":"        {"},{"line_number":691,"context_line":"            multi_io_action(m, mi, NULL, TA_SOCKET_WRITE, true);"},{"line_number":692,"context_line":"        }"}],"source_content_type":"text/x-csrc","patch_set":2,"id":"9842c1f3_cb498120","line":689,"in_reply_to":"82b89a2f_7d0af0f3","updated":"2026-11-03 10:24:51.000000000","message":"Done","commit_id":"be5e142bdaa1060f0fe42dc8f882a913d8a4e378"},{"author":{"_account_id":1000034,"name":"Gianmarco De Gregori","display_name":"Gianmarco De Gregori","email":"gianmarco@mandelbit.com","username":"its_Giaan"},"change_message_id":"39bdc5c5e4a3b71aa01fece0f899b9cf40597332","unresolved":true,"context_lines":[{"line_number":686,"context_line":"     */"},{"line_number":687,"context_line":"    {"},{"line_number":688,"context_line":"        struct multi_instance *mi;"},{"line_number":689,"context_line":"        while (!IS_SIG(\u0026m-\u003etop) \u0026\u0026 (mi \u003d multi_get_queue(m-\u003embuf)) !\u003d NULL)"},{"line_number":690,"context_line":"        {"},{"line_number":691,"context_line":"            multi_io_action(m, mi, NULL, TA_SOCKET_WRITE, true);"},{"line_number":692,"context_line":"        }"}],"source_content_type":"text/x-csrc","patch_set":2,"id":"82b89a2f_7d0af0f3","line":689,"in_reply_to":"aa7e4410_d79a5417","updated":"2026-11-03 10:06:57.000000000","message":"Well in the final series it can\u0027t: m-\u003epending is gone, so the\npacket extracted by the drain is always the one written next. In this\npatch it could, but only if m-\u003epending still holds output at the drain,\nwhich only multi_process_file_closed() can leave behind, we probably might want to get that one in first?","commit_id":"be5e142bdaa1060f0fe42dc8f882a913d8a4e378"}]}
