Initialize TLS parameters for __msan_memset() and friends in the compiler #63
Labels
Comments
ramosian-glider
pushed a commit
that referenced
this issue
Nov 21, 2022
The coreboot_table driver registers a coreboot bus while probing a "coreboot_table" device representing the coreboot table memory region. Probing this device (i.e., registering the bus) is a dependency for the module_init() functions of any driver for this bus (e.g., memconsole-coreboot.c / memconsole_driver_init()). With synchronous probe, this dependency works OK, as the link order in the Makefile ensures coreboot_table_driver_init() (and thus, coreboot_table_probe()) completes before a coreboot device driver tries to add itself to the bus. With asynchronous probe, however, coreboot_table_probe() may race with memconsole_driver_init(), and so we're liable to hit one of these two: 1. coreboot_driver_register() eventually hits "[...] the bus was not initialized.", and the memconsole driver fails to register; or 2. coreboot_driver_register() gets past #1, but still races with bus_register() and hits some other undefined/crashing behavior (e.g., in driver_find() [1]) We can resolve this by registering the bus in our initcall, and only deferring "device" work (scanning the coreboot memory region and creating sub-devices) to probe(). [1] Example failure, using 'driver_async_probe=*' kernel command line: [ 0.114217] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000010 ... [ 0.114307] CPU: 1 PID: 1 Comm: swapper/0 Not tainted 6.1.0-rc1 #63 [ 0.114316] Hardware name: Google Scarlet (DT) ... [ 0.114488] Call trace: [ 0.114494] _raw_spin_lock+0x34/0x60 [ 0.114502] kset_find_obj+0x28/0x84 [ 0.114511] driver_find+0x30/0x50 [ 0.114520] driver_register+0x64/0x10c [ 0.114528] coreboot_driver_register+0x30/0x3c [ 0.114540] memconsole_driver_init+0x24/0x30 [ 0.114550] do_one_initcall+0x154/0x2e0 [ 0.114560] do_initcall_level+0x134/0x160 [ 0.114571] do_initcalls+0x60/0xa0 [ 0.114579] do_basic_setup+0x28/0x34 [ 0.114588] kernel_init_freeable+0xf8/0x150 [ 0.114596] kernel_init+0x2c/0x12c [ 0.114607] ret_from_fork+0x10/0x20 [ 0.114624] Code: 5280002b 1100054a b900092a f9800011 (885ffc01) [ 0.114631] ---[ end trace 0000000000000000 ]--- Fixes: b81e314 ("firmware: coreboot: Make bus registration symmetric") Cc: <[email protected]> Signed-off-by: Brian Norris <[email protected]> Reviewed-by: Guenter Roeck <[email protected]> Reviewed-by: Stephen Boyd <[email protected]> Link: https://lore.kernel.org/r/20221019180934.1.If29e167d8a4771b0bf4a39c89c6946ed764817b9@changeid Signed-off-by: Greg Kroah-Hartman <[email protected]>
ramosian-glider
pushed a commit
that referenced
this issue
Nov 28, 2022
If a socket bound to a wildcard address fails to connect(), we only reset saddr and keep the port. Then, we have to fix up the bhash2 bucket; otherwise, the bucket has an inconsistent address in the list. Also, listen() for such a socket will fire the WARN_ON() in inet_csk_get_port(). [0] Note that when a system runs out of memory, we give up fixing the bucket and unlink sk from bhash and bhash2 by inet_put_port(). [0]: WARNING: CPU: 0 PID: 207 at net/ipv4/inet_connection_sock.c:548 inet_csk_get_port (net/ipv4/inet_connection_sock.c:548 (discriminator 1)) Modules linked in: CPU: 0 PID: 207 Comm: bhash2_prev_rep Not tainted 6.1.0-rc3-00799-gc8421681c845 #63 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-1.amzn2022.0.1 04/01/2014 RIP: 0010:inet_csk_get_port (net/ipv4/inet_connection_sock.c:548 (discriminator 1)) Code: 74 a7 eb 93 48 8b 54 24 18 0f b7 cb 4c 89 e6 4c 89 ff e8 48 b2 ff ff 49 8b 87 18 04 00 00 e9 32 ff ff ff 0f 0b e9 34 ff ff ff <0f> 0b e9 42 ff ff ff 41 8b 7f 50 41 8b 4f 54 89 fe 81 f6 00 00 ff RSP: 0018:ffffc900003d7e50 EFLAGS: 00010202 RAX: ffff8881047fb500 RBX: 0000000000004e20 RCX: 0000000000000000 RDX: 000000000000000a RSI: 00000000fffffe00 RDI: 00000000ffffffff RBP: ffffffff8324dc00 R08: 0000000000000001 R09: 0000000000000001 R10: 0000000000000001 R11: 0000000000000001 R12: 0000000000000000 R13: 0000000000000001 R14: 0000000000004e20 R15: ffff8881054e1280 FS: 00007f8ac04dc740(0000) GS:ffff88842fc00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000020001540 CR3: 00000001055fa003 CR4: 0000000000770ef0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 PKRU: 55555554 Call Trace: <TASK> inet_csk_listen_start (net/ipv4/inet_connection_sock.c:1205) inet_listen (net/ipv4/af_inet.c:228) __sys_listen (net/socket.c:1810) __x64_sys_listen (net/socket.c:1819 net/socket.c:1817 net/socket.c:1817) do_syscall_64 (arch/x86/entry/common.c:50 arch/x86/entry/common.c:80) entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:120) RIP: 0033:0x7f8ac051de5d Code: ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 93 af 1b 00 f7 d8 64 89 01 48 RSP: 002b:00007ffc1c177248 EFLAGS: 00000206 ORIG_RAX: 0000000000000032 RAX: ffffffffffffffda RBX: 0000000020001550 RCX: 00007f8ac051de5d RDX: ffffffffffffff80 RSI: 0000000000000000 RDI: 0000000000000004 RBP: 00007ffc1c177270 R08: 0000000000000018 R09: 0000000000000007 R10: 0000000020001540 R11: 0000000000000206 R12: 00007ffc1c177388 R13: 0000000000401169 R14: 0000000000403e18 R15: 00007f8ac0723000 </TASK> Fixes: 28044fc ("net: Add a bhash2 table hashed by port and address") Reported-by: syzbot <[email protected]> Reported-by: Mat Martineau <[email protected]> Signed-off-by: Kuniyuki Iwashima <[email protected]> Acked-by: Joanne Koong <[email protected]> Reviewed-by: Eric Dumazet <[email protected]> Signed-off-by: Jakub Kicinski <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
It may make sense to emit stores to param_tls and param_origin_tls directly in the compiler for calls to __msan_memset() for KMSAN, so that it's possible to e.g. propagate the state of uninitialized consant byte or report errors on uninitialized pointer/size arguments.
This cannot be done for MSan, because __msan_memset may be called from the userspace RTL.
The text was updated successfully, but these errors were encountered: