"value":"In the Linux kernel, the following vulnerability has been resolved:\n\nRDMA/cma: Do not change route.addr.src_addr outside state checks\n\nIf the state is not idle then resolve_prepare_src() should immediately\nfail and no change to global state should happen. However, it\nunconditionally overwrites the src_addr trying to build a temporary any\naddress.\n\nFor instance if the state is already RDMA_CM_LISTEN then this will corrupt\nthe src_addr and would cause the test in cma_cancel_operation():\n\n if (cma_any_addr(cma_src_addr(id_priv)) && !id_priv->cma_dev)\n\nWhich would manifest as this trace from syzkaller:\n\n BUG: KASAN: use-after-free in __list_add_valid+0x93/0xa0 lib/list_debug.c:26\n Read of size 8 at addr ffff8881546491e0 by task syz-executor.1/32204\n\n CPU: 1 PID: 32204 Comm: syz-executor.1 Not tainted 5.12.0-rc8-syzkaller #0\n Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011\n Call Trace:\n __dump_stack lib/dump_stack.c:79 [inline]\n dump_stack+0x141/0x1d7 lib/dump_stack.c:120\n print_address_description.constprop.0.cold+0x5b/0x2f8 mm/kasan/report.c:232\n __kasan_report mm/kasan/report.c:399 [inline]\n kasan_report.cold+0x7c/0xd8 mm/kasan/report.c:416\n __list_add_valid+0x93/0xa0 lib/list_debug.c:26\n __list_add include/linux/list.h:67 [inline]\n list_add_tail include/linux/list.h:100 [inline]\n cma_listen_on_all drivers/infiniband/core/cma.c:2557 [inline]\n rdma_listen+0x787/0xe00 drivers/infiniband/core/cma.c:3751\n ucma_listen+0x16a/0x210 drivers/infiniband/core/ucma.c:1102\n ucma_write+0x259/0x350 drivers/infiniband/core/ucma.c:1732\n vfs_write+0x28e/0xa30 fs/read_write.c:603\n ksys_write+0x1ee/0x250 fs/read_write.c:658\n do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46\n entry_SYSCALL_64_after_hwframe+0x44/0xae\n\nThis is indicating that an rdma_id_private was destroyed without doing\ncma_cancel_listens().\n\nInstead of trying to re-use the src_addr memory to indirectly create an\nany address derived from the dst build one explicitly on the stack and\nbind to that as any other normal flow would do. rdma_bind_addr() will copy\nit over the src_addr once it knows the state is valid.\n\nThis is similar to commit bc0bdc5afaa7 (\"RDMA/cma: Do not change\nroute.addr.src_addr.ss_family\")"
"value":"En el kernel de Linux, se resolvi\u00f3 la siguiente vulnerabilidad: RDMA/cma: no cambie route.addr.src_addr fuera de las comprobaciones de estado. Si el estado no est\u00e1 inactivo, resolve_prepare_src() deber\u00eda fallar inmediatamente y no deber\u00eda ocurrir ning\u00fan cambio en el estado global. Sin embargo, sobrescribe incondicionalmente src_addr al intentar crear una direcci\u00f3n temporal. Por ejemplo, si el estado ya es RDMA_CM_LISTEN, esto da\u00f1ar\u00e1 src_addr y provocar\u00e1 la prueba en cma_cancel_operation(): if (cma_any_addr(cma_src_addr(id_priv)) && !id_priv->cma_dev) Lo que se manifestar\u00eda como este rastro de syzkaller: ERROR : KASAN: use-after-free en __list_add_valid+0x93/0xa0 lib/list_debug.c:26 Lectura de tama\u00f1o 8 en addr ffff8881546491e0 por tarea syz-executor.1/32204 CPU: 1 PID: 32204 Comm: syz-executor.1 No contaminado 5.12.0-rc8-syzkaller #0 Nombre del hardware: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Seguimiento de llamadas: __dump_stack lib/dump_stack.c:79 [en l\u00ednea] dump_stack+0x141/0x1d7 lib /dump_stack.c:120 print_address_description.constprop.0.cold+0x5b/0x2f8 mm/kasan/report.c:232 __kasan_report mm/kasan/report.c:399 [en l\u00ednea] kasan_report.cold+0x7c/0xd8 mm/kasan/ report.c:416 __list_add_valid+0x93/0xa0 lib/list_debug.c:26 __list_add include/linux/list.h:67 [en l\u00ednea] list_add_tail include/linux/list.h:100 [en l\u00ednea] cma_listen_on_all drivers/infiniband/core/ cma.c:2557 [en l\u00ednea] rdma_listen+0x787/0xe00 controladores/infiniband/core/cma.c:3751 ucma_listen+0x16a/0x210 controladores/infiniband/core/ucma.c:1102 ucma_write+0x259/0x350 controladores/infiniband/core /ucma.c:1732 vfs_write+0x28e/0xa30 fs/read_write.c:603 ksys_write+0x1ee/0x250 fs/read_write.c:658 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46 Entry_SYSCALL_64_after_hwframe+0x44/ 0xae Esto indica que un rdma_id_private fue destruido sin realizar cma_cancel_listens(). En lugar de intentar reutilizar la memoria src_addr para crear indirectamente cualquier direcci\u00f3n derivada del dst, cree una expl\u00edcitamente en la pila y vinc\u00falela como lo har\u00eda cualquier otro flujo normal. rdma_bind_addr() lo copiar\u00e1 sobre src_addr una vez que sepa que el estado es v\u00e1lido. Esto es similar al commit bc0bdc5afaa7 (\"RDMA/cma: No cambiar route.addr.src_addr.ss_family\")"