Skip to contentSkip to conversation

iomi-si UBSAN warning and NULL pointer dereference

PATCH In review Meelis Roos 3 participants 6 messages Jun 19, 2018 – Jun 21, 2018 IPMI

lore
Share & tools
lore
#0 lore raw

I tried 4.18.0-rc1-00043-gba4dbdedd3ed on HP Proliant Microserver N36L and got the follsing UBSAN warning + NULL pointer dereferences. It was working without any warnings in 4.17.0.

[    7.587532] ipmi message handler version 39.2
[    7.594899] ipmi device interface
[    7.605792] IPMI System Interface driver.
[    7.605949] ipmi_si dmi-ipmi-si.0: ipmi_platform: probing via SMBIOS
[    7.606047] ipmi_si: SMBIOS: mem 0x0 regsize 1 spacing 1 irq 0
[    7.606120] ipmi_si: Adding SMBIOS-specified kcs state machine
[    7.606326] ipmi_si: Trying SMBIOS-specified kcs state machine at mem address 0x0, slave address 0x20, irq 0
[    7.606463] ipmi_si dmi-ipmi-si.0: Could not set up I/O space
[    7.606534] ================================================================================
[    7.606629] UBSAN: Undefined behaviour in drivers/char/ipmi/ipmi_msghandler.c:3477:6
[    7.606722] member access within null pointer of type 'struct ipmi_smi'
[    7.606797] CPU: 1 PID: 1360 Comm: systemd-udevd Not tainted 4.18.0-rc1-00043-gba4dbdedd3ed #26
[    7.606892] Hardware name: HP ProLiant MicroServer, BIOS O41     10/01/2013
[    7.606962] Call Trace:
[    7.607042]  ? dump_stack+0x5a/0x9b
[    7.607116]  ? ubsan_epilogue+0x9/0x40
[    7.607188]  ? ubsan_type_mismatch_common+0x11f/0x1a0
[    7.607260]  ? __ubsan_handle_type_mismatch+0x3a/0x60
[    7.607337]  ? ipmi_unregister_smi+0x55c/0x570 [ipmi_msghandler]
[    7.607424]  ? try_smi_init+0xbaa/0x1ab5 [ipmi_si]
[    7.607509]  ? init_ipmi_si+0x158/0x240 [ipmi_si]
[    7.607590]  ? ipmi_si_add_smi+0x390/0x390 [ipmi_si]
[    7.607662]  ? do_one_initcall+0x58/0x230
[    7.607735]  ? kmem_cache_alloc+0x43/0x1f0
[    7.607807]  ? do_init_module+0xa7/0x2a9
[    7.607877]  ? load_module+0x1f40/0x3510
[    7.607947]  ? __symbol_put+0x80/0x80
[    7.608020]  ? kernel_read_file+0x229/0x3a0
[    7.608092]  ? __do_sys_finit_module+0xfa/0x120
[    7.608163]  ? do_syscall_64+0x5a/0x1e0
[    7.608233]  ? page_fault+0x8/0x30
[    7.608306]  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9
[    7.608376] ================================================================================
[    7.608503] BUG: unable to handle kernel NULL pointer dereference at 0000000000000000
[    7.608600] PGD 0 P4D 0 
[    7.608672] Oops: 0000 [#1] SMP NOPTI
[    7.608743] CPU: 1 PID: 1360 Comm: systemd-udevd Not tainted 4.18.0-rc1-00043-gba4dbdedd3ed #26
[    7.608836] Hardware name: HP ProLiant MicroServer, BIOS O41     10/01/2013
[    7.608913] RIP: 0010:ipmi_unregister_smi+0x31/0x570 [ipmi_msghandler]
[    7.608982] Code: 54 55 48 89 fd 53 48 83 ec 30 65 48 8b 04 25 28 00 00 00 48 89 44 24 28 31 c0 48 85 ff 0f 84 24 05 00 00 48 c7 c7 c0 23 16 c0 <44> 8b 65 00 e8 a6 65 5c c2 48 83 fd f0 c7 45 00 ff ff ff ff c6 45 
[    7.609210] RSP: 0018:ffffa52c40227bb8 EFLAGS: 00010292
[    7.609281] RAX: 0000000000000000 RBX: ffff8e8e3b2df200 RCX: 0000000000000006
[    7.609352] RDX: 0000000000000000 RSI: 0000000000000202 RDI: ffffffffc01623c0
[    7.609424] RBP: 0000000000000000 R08: 0000000000000199 R09: 000000000000025a
[    7.609495] R10: ffffffff821bc0b0 R11: 0000000000000006 R12: ffffffffc0181aa8
[    7.609566] R13: 0000000000000000 R14: ffff8e8e3b2df240 R15: ffffffffc0181260
[    7.609640] FS:  00007fef3a80b8c0(0000) GS:ffff8e8e3dd00000(0000) knlGS:0000000000000000
[    7.609734] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    7.609803] CR2: 0000000000000000 CR3: 000000003ab1a000 CR4: 00000000000006e0
[    7.609873] Call Trace:
[    7.609956]  ? try_smi_init+0xbaa/0x1ab5 [ipmi_si]
[    7.610040]  ? init_ipmi_si+0x158/0x240 [ipmi_si]
[    7.610121]  ? ipmi_si_add_smi+0x390/0x390 [ipmi_si]
[    7.610191]  ? do_one_initcall+0x58/0x230
[    7.610262]  ? kmem_cache_alloc+0x43/0x1f0
[    7.610333]  ? do_init_module+0xa7/0x2a9
[    7.610404]  ? load_module+0x1f40/0x3510
[    7.610475]  ? __symbol_put+0x80/0x80
[    7.610547]  ? kernel_read_file+0x229/0x3a0
[    7.610618]  ? __do_sys_finit_module+0xfa/0x120
[    7.610689]  ? do_syscall_64+0x5a/0x1e0
[    7.610759]  ? page_fault+0x8/0x30
[    7.610832]  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9
[    7.610902] Modules linked in: ipmi_si(+) ipmi_devintf ipmi_msghandler k10temp jc42 w83795 eeprom ip_tables
[    7.611014] CR2: 0000000000000000
[    7.611094] ---[ end trace 099b4ef2a90b74a1 ]---
[    7.611170] RIP: 0010:ipmi_unregister_smi+0x31/0x570 [ipmi_msghandler]
[    7.611239] Code: 54 55 48 89 fd 53 48 83 ec 30 65 48 8b 04 25 28 00 00 00 48 89 44 24 28 31 c0 48 85 ff 0f 84 24 05 00 00 48 c7 c7 c0 23 16 c0 <44> 8b 65 00 e8 a6 65 5c c2 48 83 fd f0 c7 45 00 ff ff ff ff c6 45 
[    7.611466] RSP: 0018:ffffa52c40227bb8 EFLAGS: 00010292
[    7.611537] RAX: 0000000000000000 RBX: ffff8e8e3b2df200 RCX: 0000000000000006
[    7.611609] RDX: 0000000000000000 RSI: 0000000000000202 RDI: ffffffffc01623c0
[    7.611680] RBP: 0000000000000000 R08: 0000000000000199 R09: 000000000000025a
[    7.611751] R10: ffffffff821bc0b0 R11: 0000000000000006 R12: ffffffffc0181aa8
[    7.611822] R13: 0000000000000000 R14: ffff8e8e3b2df240 R15: ffffffffc0181260
[    7.611894] FS:  00007fef3a80b8c0(0000) GS:ffff8e8e3dd00000(0000) knlGS:0000000000000000
[    7.611988] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    7.612067] CR2: 0000000000000000 CR3: 000000003ab1a000 CR4: 00000000000006e0
Signature · 1 line
Meelis Roos ([email protected])
#1 lore raw

PATCH 1/1 ipmi: Cleanup oops on initialization failure

From: Corey Minyard <[email protected]>

Commit 93c303d2045b3 "ipmi_si: Clean up shutdown a bit" didn't copy the behavior of the cleanup in one spot, it needed to check for a non-NULL interface before cleaning it up.

---

This patch should fix the issue.

BTW, can you send me at least the IPMI portion of the output of dmidecode for your machine? I have seen a lot of these where the address in the SMBIOS tables is incorrect, and I'm wondering if it's something in the driver, or if it's really the tables that are bad.

Thanks for reporting this. On your tested-by I'll send this up to Linus.

-corey

 drivers/char/ipmi/ipmi_si_intf.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/char/ipmi/ipmi_si_intf.c +4 -2
index 3d0add6..a5987f8 100644
@@ -2088,8 +2088,10 @@ static int try_smi_init(struct smi_info *new_smi)
 	return 0;
 
 out_err:
-	ipmi_unregister_smi(new_smi->intf);
-	new_smi->intf = NULL;
+	if (new_smi->intf) {
+		ipmi_unregister_smi(new_smi->intf);
+		new_smi->intf = NULL;
+	}
 
 	kfree(init_name);
 
Signature · 1 line
2.7.4
author Tested-by on v1 1/1
#2 lore raw

Re: [PATCH] ipmi: Cleanup oops on initialization failure

Quoted 5 lines

Commit 93c303d2045b3 "ipmi_si: Clean up shutdown a bit" didn't copy the behavior of the cleanup in one spot, it needed to check for a non-NULL interface before cleaning it up.

The corresponding dmesg:

[    7.372830] IPMI System Interface driver.
[    7.373034] ipmi_si dmi-ipmi-si.0: ipmi_platform: probing via SMBIOS
[    7.373109] ipmi_si: SMBIOS: mem 0x0 regsize 1 spacing 1 irq 0
[    7.373182] ipmi_si: Adding SMBIOS-specified kcs state machine
[    7.373352] ipmi_si: Trying SMBIOS-specified kcs state machine at mem address 0x0, slave address 0x20, irq 0
[    7.373479] ipmi_si dmi-ipmi-si.0: Could not set up I/O space
Quoted 5 lines

BTW, can you send me at least the IPMI portion of the output of dmidecode for your machine? I have seen a lot of these where the address in the SMBIOS tables is incorrect, and I'm wondering if it's something in the driver, or if it's really the tables that are bad.

Handle 0x001B, DMI type 38, 18 bytes
IPMI Device Information
Interface Type: KCS (Keyboard Control Style) Specification Version: 2.0
I2C Slave Address: 0x10
NV Storage Device: Not Present
Base Address: 0x0000000000000000 (Memory-mapped) Register Spacing: Successive Byte Boundaries

Quoted 27 lines

Thanks for reporting this. On your tested-by I'll send this up to Linus.

(quoted diff) +0 -1
-corey
 
 drivers/char/ipmi/ipmi_si_intf.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/char/ipmi/ipmi_si_intf.c +4 -2
index 3d0add6..a5987f8 100644
@@ -2088,8 +2088,10 @@ static int try_smi_init(struct smi_info *new_smi)
 	return 0;
 
 out_err:
-	ipmi_unregister_smi(new_smi->intf);
-	new_smi->intf = NULL;
+	if (new_smi->intf) {
+		ipmi_unregister_smi(new_smi->intf);
+		new_smi->intf = NULL;
+	}
 
 	kfree(init_name);
 
Signature · 1 line
Meelis Roos ([email protected])
#3 lore raw

Re: [PATCH] ipmi: Cleanup oops on initialization failure

Quoted 22 lines · Meelis Roos
Quoted 5 lines

Commit 93c303d2045b3 "ipmi_si: Clean up shutdown a bit" didn't copy the behavior of the cleanup in one spot, it needed to check for a non-NULL interface before cleaning it up.

The corresponding dmesg:

[    7.372830] IPMI System Interface driver.
[    7.373034] ipmi_si dmi-ipmi-si.0: ipmi_platform: probing via SMBIOS
[    7.373109] ipmi_si: SMBIOS: mem 0x0 regsize 1 spacing 1 irq 0
[    7.373182] ipmi_si: Adding SMBIOS-specified kcs state machine
[    7.373352] ipmi_si: Trying SMBIOS-specified kcs state machine at mem address 0x0, slave address 0x20, irq 0
[    7.373479] ipmi_si dmi-ipmi-si.0: Could not set up I/O space
Quoted 5 lines

BTW, can you send me at least the IPMI portion of the output of dmidecode for your machine? I have seen a lot of these where the address in the SMBIOS tables is incorrect, and I'm wondering if it's something in the driver, or if it's really the tables that are bad.

Quoted 8 lines

Handle 0x001B, DMI type 38, 18 bytes
IPMI Device Information
Interface Type: KCS (Keyboard Control Style) Specification Version: 2.0
I2C Slave Address: 0x10
NV Storage Device: Not Present
Base Address: 0x0000000000000000 (Memory-mapped) Register Spacing: Successive Byte Boundaries

Thanks a bunch.  It looks like the SMBIOS tables are wrong.  I wonder if this is what some vendor do if there is no IPMI device installed.  I guess I need to add a check for this.

-corey

Quoted 26 lines
Quoted 26 lines

Thanks for reporting this. On your tested-by I'll send this up to Linus.

(quoted diff) +0 -1
-corey
 
  drivers/char/ipmi/ipmi_si_intf.c | 6 ++++--
  1 file changed, 4 insertions(+), 2 deletions(-)
drivers/char/ipmi/ipmi_si_intf.c +4 -2
index 3d0add6..a5987f8 100644
@@ -2088,8 +2088,10 @@ static int try_smi_init(struct smi_info *new_smi)
  	return 0;
 
  out_err:
-	ipmi_unregister_smi(new_smi->intf);
-	new_smi->intf = NULL;
+	if (new_smi->intf) {
+		ipmi_unregister_smi(new_smi->intf);
+		new_smi->intf = NULL;
+	}
 
  	kfree(init_name);
 
#4 lore raw

Re: [PATCH] ipmi: Cleanup oops on initialization failure

Quoted 24 lines
Quoted 15 lines

The corresponding dmesg:

[    7.372830] IPMI System Interface driver.
[    7.373034] ipmi_si dmi-ipmi-si.0: ipmi_platform: probing via SMBIOS
[    7.373109] ipmi_si: SMBIOS: mem 0x0 regsize 1 spacing 1 irq 0
[    7.373182] ipmi_si: Adding SMBIOS-specified kcs state machine
[    7.373352] ipmi_si: Trying SMBIOS-specified kcs state machine at mem
address 0x0, slave address 0x20, irq 0
[    7.373479] ipmi_si dmi-ipmi-si.0: Could not set up I/O space
Quoted 5 lines

BTW, can you send me at least the IPMI portion of the output of dmidecode for your machine? I have seen a lot of these where the address in the SMBIOS tables is incorrect, and I'm wondering if it's something in the driver, or if it's really the tables that are bad.

Quoted 8 lines

Handle 0x001B, DMI type 38, 18 bytes
IPMI Device Information
Interface Type: KCS (Keyboard Control Style) Specification Version: 2.0
I2C Slave Address: 0x10
NV Storage Device: Not Present
Base Address: 0x0000000000000000 (Memory-mapped) Register Spacing: Successive Byte Boundaries

Quoted 3 lines

Thanks a bunch.  It looks like the SMBIOS tables are wrong.  I wonder if this is what some vendor do if there is no IPMI device installed.  I guess I need to add a check for this.

Another machine (Sun X2100) with similar crash is also cured by the patch, but this is slightly different (not NULL):

[    8.891217] IPMI System Interface driver.
[    8.898404] ipmi_si dmi-ipmi-si.0: ipmi_platform: probing via SMBIOS
[    8.905635] ipmi_si: SMBIOS: io 0xca2 regsize 1 spacing 1 irq 0
[    8.912895] ipmi_si: Adding SMBIOS-specified kcs state machine
[    8.920246] ipmi_si: Trying SMBIOS-specified kcs state machine at i/o address 0xca2, slave address 0x20, irq 0
[    8.934379] ipmi_si dmi-ipmi-si.0: Interface detection failed

IPMI Device Information
Interface Type: KCS (Keyboard Control Style) Specification Version: 1.5
I2C Slave Address: 0x10
NV Storage Device: Not Present
Base Address: 0x0000000000000CA2 (I/O) Register Spacing: Successive Byte Boundaries

Signature · 1 line
Meelis Roos ([email protected])
#5 lore raw

Re: [PATCH] ipmi: Cleanup oops on initialization failure

Quoted 36 lines · Meelis Roos
Quoted 23 lines
Quoted 15 lines

The corresponding dmesg:

[    7.372830] IPMI System Interface driver.
[    7.373034] ipmi_si dmi-ipmi-si.0: ipmi_platform: probing via SMBIOS
[    7.373109] ipmi_si: SMBIOS: mem 0x0 regsize 1 spacing 1 irq 0
[    7.373182] ipmi_si: Adding SMBIOS-specified kcs state machine
[    7.373352] ipmi_si: Trying SMBIOS-specified kcs state machine at mem
address 0x0, slave address 0x20, irq 0
[    7.373479] ipmi_si dmi-ipmi-si.0: Could not set up I/O space
Quoted 5 lines

BTW, can you send me at least the IPMI portion of the output of dmidecode for your machine? I have seen a lot of these where the address in the SMBIOS tables is incorrect, and I'm wondering if it's something in the driver, or if it's really the tables that are bad.

Quoted 8 lines

Handle 0x001B, DMI type 38, 18 bytes
IPMI Device Information
Interface Type: KCS (Keyboard Control Style) Specification Version: 2.0
I2C Slave Address: 0x10
NV Storage Device: Not Present
Base Address: 0x0000000000000000 (Memory-mapped) Register Spacing: Successive Byte Boundaries

Quoted 3 lines

Thanks a bunch.  It looks like the SMBIOS tables are wrong.  I wonder if this is what some vendor do if there is no IPMI device installed.  I guess I need to add a check for this.

Another machine (Sun X2100) with similar crash is also cured by the patch, but this is slightly different (not NULL):

[    8.891217] IPMI System Interface driver.
[    8.898404] ipmi_si dmi-ipmi-si.0: ipmi_platform: probing via SMBIOS
[    8.905635] ipmi_si: SMBIOS: io 0xca2 regsize 1 spacing 1 irq 0
[    8.912895] ipmi_si: Adding SMBIOS-specified kcs state machine
[    8.920246] ipmi_si: Trying SMBIOS-specified kcs state machine at i/o address 0xca2, slave address 0x20, irq 0
[    8.934379] ipmi_si dmi-ipmi-si.0: Interface detection failed
Quoted 8 lines

IPMI Device Information
Interface Type: KCS (Keyboard Control Style) Specification Version: 1.5
I2C Slave Address: 0x10
NV Storage Device: Not Present
Base Address: 0x0000000000000CA2 (I/O) Register Spacing: Successive Byte Boundaries

That's even worse.  The SMBIOS table says the interface is there, but it's not there.  Not much I can do about that :(.

Thanks again,

-corey