"value":"In the Linux kernel, the following vulnerability has been resolved:\n\nbonding: change ipsec_lock from spin lock to mutex\n\nIn the cited commit, bond->ipsec_lock is added to protect ipsec_list,\nhence xdo_dev_state_add and xdo_dev_state_delete are called inside\nthis lock. As ipsec_lock is a spin lock and such xfrmdev ops may sleep,\n\"scheduling while atomic\" will be triggered when changing bond's\nactive slave.\n\n[ 101.055189] BUG: scheduling while atomic: bash/902/0x00000200\n[ 101.055726] Modules linked in:\n[ 101.058211] CPU: 3 PID: 902 Comm: bash Not tainted 6.9.0-rc4+ #1\n[ 101.058760] Hardware name:\n[ 101.059434] Call Trace:\n[ 101.059436] <TASK>\n[ 101.060873] dump_stack_lvl+0x51/0x60\n[ 101.061275] __schedule_bug+0x4e/0x60\n[ 101.061682] __schedule+0x612/0x7c0\n[ 101.062078] ? __mod_timer+0x25c/0x370\n[ 101.062486] schedule+0x25/0xd0\n[ 101.062845] schedule_timeout+0x77/0xf0\n[ 101.063265] ? asm_common_interrupt+0x22/0x40\n[ 101.063724] ? __bpf_trace_itimer_state+0x10/0x10\n[ 101.064215] __wait_for_common+0x87/0x190\n[ 101.064648] ? usleep_range_state+0x90/0x90\n[ 101.065091] cmd_exec+0x437/0xb20 [mlx5_core]\n[ 101.065569] mlx5_cmd_do+0x1e/0x40 [mlx5_core]\n[ 101.066051] mlx5_cmd_exec+0x18/0x30 [mlx5_core]\n[ 101.066552] mlx5_crypto_create_dek_key+0xea/0x120 [mlx5_core]\n[ 101.067163] ? bonding_sysfs_store_option+0x4d/0x80 [bonding]\n[ 101.067738] ? kmalloc_trace+0x4d/0x350\n[ 101.068156] mlx5_ipsec_create_sa_ctx+0x33/0x100 [mlx5_core]\n[ 101.068747] mlx5e_xfrm_add_state+0x47b/0xaa0 [mlx5_core]\n[ 101.069312] bond_change_active_slave+0x392/0x900 [bonding]\n[ 101.069868] bond_option_active_slave_set+0x1c2/0x240 [bonding]\n[ 101.070454] __bond_opt_set+0xa6/0x430 [bonding]\n[ 101.070935] __bond_opt_set_notify+0x2f/0x90 [bonding]\n[ 101.071453] bond_opt_tryset_rtnl+0x72/0xb0 [bonding]\n[ 101.071965] bonding_sysfs_store_option+0x4d/0x80 [bonding]\n[ 101.072567] kernfs_fop_write_iter+0x10c/0x1a0\n[ 101.073033] vfs_write+0x2d8/0x400\n[ 101.073416] ? alloc_fd+0x48/0x180\n[ 101.073798] ksys_write+0x5f/0xe0\n[ 101.074175] do_syscall_64+0x52/0x110\n[ 101.074576] entry_SYSCALL_64_after_hwframe+0x4b/0x53\n\nAs bond_ipsec_add_sa_all and bond_ipsec_del_sa_all are only called\nfrom bond_change_active_slave, which requires holding the RTNL lock.\nAnd bond_ipsec_add_sa and bond_ipsec_del_sa are xfrm state\nxdo_dev_state_add and xdo_dev_state_delete APIs, which are in user\ncontext. So ipsec_lock doesn't have to be spin lock, change it to\nmutex, and thus the above issue can be resolved."
"value":"En el kernel de Linux, se ha resuelto la siguiente vulnerabilidad: bonding: cambiar ipsec_lock de spin lock a mutex en el commit citado, se agrega bond->ipsec_lock para proteger ipsec_list, por lo tanto, se llaman xdo_dev_state_add y xdo_dev_state_delete dentro de este bloqueo. Como ipsec_lock es un spin lock y tales operaciones xfrmdev pueden dormir, se activar\u00e1 \"programaci\u00f3n mientras es at\u00f3mica\" al cambiar el esclavo activo de bond. [ 101.055189] ERROR: programaci\u00f3n mientras es at\u00f3mica: bash/902/0x00000200 [ 101.055726] M\u00f3dulos vinculados en: [ 101.058211] CPU: 3 PID: 902 Comm: bash No contaminado 6.9.0-rc4+ #1 [ 101.058760] Nombre del hardware: [ 101.059434] Seguimiento de llamadas: [ 101.059436] [ 101.060873] dump_stack_lvl+0x51/0x60 [ 101.061275] __schedule_bug+0x4e/0x60 [ 101.061682] __schedule+0x612/0x7c0 [ 101.062078] ? __mod_timer+0x25c/0x370 [ 101.062486] schedule+0x25/0xd0 [ 101.062845] schedule_timeout+0x77/0xf0 [ 101.063265] ? asm_common_interrupt+0x22/0x40 [ 101.063724] ? __bpf_trace_itimer_state+0x10/0x10 [ 101.064215] __wait_for_common+0x87/0x190 [ 101.064648] ? opci\u00f3n_almacenamiento_sysfs_bonding+0x4d/0x80 [bonding] [ 101.067738] ? kmalloc_trace+0x4d/0x350 [ 101.068156] mlx5_ipsec_create_sa_ctx+0x33/0x100 [mlx5_core] [ 101.068747] mlx5e_xfrm_add_state+0x47b/0xaa0 [mlx5_core] [ 101.069312] cambio_enlace_esclavo_activo+0x392/0x900 [enlace] [ 101.069868] opci\u00f3n_enlace_esclavo_activo_conjunto+0x1c2/0x240 [enlace] [ 101.070454] __opci\u00f3n_enlace_conjunto+0xa6/0x430 [enlace] [ 101.070935] __bond_opt_set_notify+0x2f/0x90 [vinculaci\u00f3n] [ 101.071453] bond_opt_tryset_rtnl+0x72/0xb0 [vinculaci\u00f3n] [ 101.071965] bonding_sysfs_store_option+0x4d/0x80 [vinculaci\u00f3n] [ 101.072567] kernfs_fop_write_iter+0x10c/0x1a0 [ 101.073033] vfs_write+0x2d8/0x400 [ 101.073416] ? alloc_fd+0x48/0x180 [ 101.073798] ksys_write+0x5f/0xe0 [ 101.074175] do_syscall_64+0x52/0x110 [ 101.074576] entry_SYSCALL_64_after_hwframe+0x4b/0x53 Como bond_ipsec_add_sa_all y bond_ipsec_del_sa_all solo se llaman desde bond_change_active_slave, que requiere mantener el bloqueo RTNL. Y bond_ipsec_add_sa y bond_ipsec_del_sa son API xdo_dev_state_add y xdo_dev_state_delete de estado xfrm, que est\u00e1n en el contexto del usuario. Por lo tanto, ipsec_lock no tiene que ser un bloqueo de giro; c\u00e1mbielo a mutex y, por lo tanto, se puede resolver el problema anterior."