Skip to contentSkip to conversation

SHM oops in newseg()

DISCUSSION Dave Jones 5 participants 5 messages May 9, 2013

Subsystems (2)
lore
Share & tools
lore

Dave Jones

author
#0 lore raw

Just saw this on v3.9-11789-ge0fd9af while fuzz-testing.

[  163.917836] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[  163.918984] IP: [<ffffffff812c48ed>] newseg+0x10d/0x390
[  163.919705] PGD cf8c1067 PUD cf8c2067 PMD 0 
[  163.920326] Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC
[  163.921096] Modules linked in: hidp bnep l2tp_ppp l2tp_netlink l2tp_core ipt_ULOG rfcomm can_bcm cmtp kernelcapi nfnetlink scsi_transport_iscsi can_raw ipx p8023 pppoe p8022 phonet af_rxrpc pppox caif_socket ppp_generic caif rose af_802154 slhc llc2 nfc appletalk decnet atm rds psnap llc can netrom af_key x25 ax25 irda crc_ccitt lockd sunrpc ip6t_REJECT nf_conntrack_ipv6 nf_defrag_ipv6 xt_conntrack nf_conntrack ip6table_filter ip6_tables snd_hda_codec_realtek snd_hda_intel raid0 snd_hda_codec kvm_amd snd_pcm kvm btusb bluetooth microcode serio_raw snd_page_alloc snd_timer pcspkr snd rfkill edac_core soundcore r8169 mii radeon backlight drm_kms_helper ttm
[  163.929949] CPU: 2 PID: 7636 Comm: trinity-child2 Not tainted 3.9.0+ #67
[  163.931276] Hardware name: Gigabyte Technology Co., Ltd. GA-MA78GM-S2H/GA-MA78GM-S2H, BIOS F12a 04/23/2010
[  163.932551] task: ffff88011bc04a00 ti: ffff8800cf920000 task.ti: ffff8800cf920000
[  163.933541] RIP: 0010:[<ffffffff812c48ed>]  [<ffffffff812c48ed>] newseg+0x10d/0x390
[  163.934573] RSP: 0018:ffff8800cf921e78  EFLAGS: 00010246
[  163.935281] RAX: 0000000000000000 RBX: ffffffff81c77a20 RCX: 0000000000000009
[  163.936225] RDX: ffffffff82ab62b8 RSI: 0000000000001000 RDI: 0000000000008000
[  163.937169] RBP: ffff8800cf921ec8 R08: 0000000000000010 R09: 000000000000000f
[  163.938113] R10: 0000000000000000 R11: 0000000000000001 R12: 0000000000001fff
[  163.939057] R13: ffff8800cfa30b80 R14: 000000003f43df5b R15: 0000000000000000
[  163.940005] FS:  00007f8391182740(0000) GS:ffff88012aa00000(0000) knlGS:0000000000000000
[  163.941077] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  163.941843] CR2: 0000000000000008 CR3: 00000000cf8c0000 CR4: 00000000000007e0
[  163.942789] Stack:
[  163.943067]  0000000f81c77c50 0000000000000002 3056535953c77c50 0030303030303030
[  163.944115]  00000000bf3b8607 0000000000000000 0000000000000010 ffffffff81c77c50
[  163.948881]  00000000000000e8 0000000000000002 ffff8800cf921f40 ffffffff812be322
[  163.953629] Call Trace:
[  163.957706]  [<ffffffff812be322>] ipcget+0x182/0x380
[  163.962123]  [<ffffffff810b99a5>] ? trace_hardirqs_on_caller+0x115/0x1e0
[  163.966752]  [<ffffffff812c559a>] SyS_shmget+0x5a/0x60
[  163.971163]  [<ffffffff812c47e0>] ? shm_close+0x140/0x140
[  163.975590]  [<ffffffff812c3e60>] ? shm_release+0x50/0x50
[  163.979991]  [<ffffffff812c3df0>] ? shm_get_unmapped_area+0x20/0x20
[  163.984499]  [<ffffffff816caa14>] tracesys+0xdd/0xe2
[  163.988868] Code: c6 00 08 00 00 0f 84 f3 00 00 00 45 89 f1 41 c1 e9 1a 0f 85 2e 01 00 00 8b 05 70 1e 7f 01 48 69 c0 98 04 00 00 48 05 20 5e ab 82 <8b> 48 08 b8 00 10 00 00 4c 89 f2 48 c1 e2 09 48 8d 7d c3 41 b8 
[  164.000083] RIP  [<ffffffff812c48ed>] newseg+0x10d/0x390
[  164.004681]  RSP <ffff8800cf921e78>
[  164.009000] CR2: 0000000000000008
[  164.064784] ---[ end trace e74858ae5e89f4a7 ]---

That Code: line decoded seems to point at..

static inline unsigned long huge_page_size(struct hstate *h)

{
        return (unsigned long)PAGE_SIZE << h->order;
     bfd:       8b 48 08                mov    0x8(%rax),%ecx
     c00:       b8 00 10 00 00          mov    $0x1000,%eax
#1 lore raw
Cc Naoya Horiguchi <[email protected]>
Quoted 4 lines · Dave Jones

Just saw this on v3.9-11789-ge0fd9af while fuzz-testing.

[  163.917836] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[  163.918984] IP: [<ffffffff812c48ed>] newseg+0x10d/0x390

The patch below should fix it.

===========================

[PATCH] shm: fix null pointer deref when userspace specifies invalid hugepage size

Dave reported an oops triggered by trinity:

[  163.917836] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[  163.918984] IP: [<ffffffff812c48ed>] newseg+0x10d/0x390
[  163.919705] PGD cf8c1067 PUD cf8c2067 PMD 0
[  163.920326] Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC
[  163.929949] CPU: 2 PID: 7636 Comm: trinity-child2 Not tainted 3.9.0+#67
...
[  163.953629] Call Trace:
[  163.957706]  [<ffffffff812be322>] ipcget+0x182/0x380
[  163.962123]  [<ffffffff810b99a5>] ?trace_hardirqs_on_caller+0x115/0x1e0
[  163.966752]  [<ffffffff812c559a>] SyS_shmget+0x5a/0x60
[  163.971163]  [<ffffffff812c47e0>] ? shm_close+0x140/0x140
[  163.975590]  [<ffffffff812c3e60>] ? shm_release+0x50/0x50
[  163.979991]  [<ffffffff812c3df0>] ? shm_get_unmapped_area+0x20/0x20
[  163.984499]  [<ffffffff816caa14>] tracesys+0xdd/0xe2

This bug was introduced by commit af73e4d9506d3b797509f3c030e7dcd554f7d9c4 ("hugetlbfs: fix mmap failure in unaligned size request").

---

Previously it would return -ENODEV, but seems -EINVAL is more appropriate.

---
 ipc/shm.c | 8 +++++++-
 mm/mmap.c | 8 ++++++--
 2 files changed, 13 insertions(+), 3 deletions(-)
ipc/shm.c +7 -1
index 34af1fe..7e199fa 100644
@@ -493,7 +493,13 @@ static int newseg(struct ipc_namespace *ns, struct ipc_params *params)
 	if (shmflg & SHM_HUGETLB) {
 		struct hstate *hs = hstate_sizelog((shmflg >> SHM_HUGE_SHIFT)
 						& SHM_HUGE_MASK);
-		size_t hugesize = ALIGN(size, huge_page_size(hs));
+		size_t hugesize;
+
+		if (!hs) {
+			error = -EINVAL;
+			goto no_file;
+		}
+		hugesize = ALIGN(size, huge_page_size(hs));
 
 		/* hugetlb_file_setup applies strict accounting */
 		if (shmflg & SHM_NORESERVE)
mm/mmap.c +6 -2
index 1ae21d6..f681e18 100644
@@ -1367,9 +1367,13 @@ SYSCALL_DEFINE6(mmap_pgoff, unsigned long, addr, unsigned long, len,
 			len = ALIGN(len, huge_page_size(hstate_file(file)));
 	} else if (flags & MAP_HUGETLB) {
 		struct user_struct *user = NULL;
+		struct hstate *hs = hstate_sizelog((flags >> MAP_HUGE_SHIFT) &
+						   SHM_HUGE_MASK);
 
-		len = ALIGN(len, huge_page_size(hstate_sizelog(
-			(flags >> MAP_HUGE_SHIFT) & MAP_HUGE_MASK)));
+		if (!hs)
+			return -EINVAL;
+
+		len = ALIGN(len, huge_page_size(hs));
 		/*
 		 * VM_NORESERVE is used because the reservations will be
 		 * taken when vm_ops->mmap() is called
Signature · 1 line
1.8.0.2
Reviewed-by
#2 lore raw
Quoted 31 lines · Li Zefan
Cc Naoya Horiguchi <[email protected]>
Quoted 4 lines · Dave Jones

Just saw this on v3.9-11789-ge0fd9af while fuzz-testing.

[  163.917836] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[  163.918984] IP: [<ffffffff812c48ed>] newseg+0x10d/0x390

The patch below should fix it.

===========================

[PATCH] shm: fix null pointer deref when userspace specifies invalid hugepage size

Dave reported an oops triggered by trinity:

[  163.917836] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[  163.918984] IP: [<ffffffff812c48ed>] newseg+0x10d/0x390
[  163.919705] PGD cf8c1067 PUD cf8c2067 PMD 0
[  163.920326] Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC
[  163.929949] CPU: 2 PID: 7636 Comm: trinity-child2 Not tainted 3.9.0+#67
...
[  163.953629] Call Trace:
[  163.957706]  [<ffffffff812be322>] ipcget+0x182/0x380
[  163.962123]  [<ffffffff810b99a5>] ?trace_hardirqs_on_caller+0x115/0x1e0
[  163.966752]  [<ffffffff812c559a>] SyS_shmget+0x5a/0x60
[  163.971163]  [<ffffffff812c47e0>] ? shm_close+0x140/0x140
[  163.975590]  [<ffffffff812c3e60>] ? shm_release+0x50/0x50
[  163.979991]  [<ffffffff812c3df0>] ? shm_get_unmapped_area+0x20/0x20
[  163.984499]  [<ffffffff816caa14>] tracesys+0xdd/0xe2
Quoted 6 lines

This bug was introduced by commit af73e4d9506d3b797509f3c030e7dcd554f7d9c4 ("hugetlbfs: fix mmap failure in unaligned size request").

Thank you.

Quoted 3 lines

---

Previously it would return -ENODEV, but seems -EINVAL is more appropriate.

Agreed.

Thanks,
Naoya

Quoted 49 lines
---
 ipc/shm.c | 8 +++++++-
 mm/mmap.c | 8 ++++++--
 2 files changed, 13 insertions(+), 3 deletions(-)
ipc/shm.c +7 -1
index 34af1fe..7e199fa 100644
@@ -493,7 +493,13 @@ static int newseg(struct ipc_namespace *ns, struct ipc_params *params)
 	if (shmflg & SHM_HUGETLB) {
 		struct hstate *hs = hstate_sizelog((shmflg >> SHM_HUGE_SHIFT)
 						& SHM_HUGE_MASK);
-		size_t hugesize = ALIGN(size, huge_page_size(hs));
+		size_t hugesize;
+
+		if (!hs) {
+			error = -EINVAL;
+			goto no_file;
+		}
+		hugesize = ALIGN(size, huge_page_size(hs));
 
 		/* hugetlb_file_setup applies strict accounting */
 		if (shmflg & SHM_NORESERVE)
mm/mmap.c +6 -2
index 1ae21d6..f681e18 100644
@@ -1367,9 +1367,13 @@ SYSCALL_DEFINE6(mmap_pgoff, unsigned long, addr, unsigned long, len,
 			len = ALIGN(len, huge_page_size(hstate_file(file)));
 	} else if (flags & MAP_HUGETLB) {
 		struct user_struct *user = NULL;
+		struct hstate *hs = hstate_sizelog((flags >> MAP_HUGE_SHIFT) &
+						   SHM_HUGE_MASK);
 
-		len = ALIGN(len, huge_page_size(hstate_sizelog(
-			(flags >> MAP_HUGE_SHIFT) & MAP_HUGE_MASK)));
+		if (!hs)
+			return -EINVAL;
+
+		len = ALIGN(len, huge_page_size(hs));
 		/*
 		 * VM_NORESERVE is used because the reservations will be
 		 * taken when vm_ops->mmap() is called
Signature · 1 line
1.8.0.2
Acked-by
#3 lore raw
In reply to Li Zefan
Quoted 31 lines · Li Zefan
Cc Naoya Horiguchi <[email protected]>
Quoted 4 lines · Dave Jones

Just saw this on v3.9-11789-ge0fd9af while fuzz-testing.

[  163.917836] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[  163.918984] IP: [<ffffffff812c48ed>] newseg+0x10d/0x390

The patch below should fix it.

===========================

[PATCH] shm: fix null pointer deref when userspace specifies invalid hugepage size

Dave reported an oops triggered by trinity:

[  163.917836] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[  163.918984] IP: [<ffffffff812c48ed>] newseg+0x10d/0x390
[  163.919705] PGD cf8c1067 PUD cf8c2067 PMD 0
[  163.920326] Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC
[  163.929949] CPU: 2 PID: 7636 Comm: trinity-child2 Not tainted 3.9.0+#67
...
[  163.953629] Call Trace:
[  163.957706]  [<ffffffff812be322>] ipcget+0x182/0x380
[  163.962123]  [<ffffffff810b99a5>] ?trace_hardirqs_on_caller+0x115/0x1e0
[  163.966752]  [<ffffffff812c559a>] SyS_shmget+0x5a/0x60
[  163.971163]  [<ffffffff812c47e0>] ? shm_close+0x140/0x140
[  163.975590]  [<ffffffff812c3e60>] ? shm_release+0x50/0x50
[  163.979991]  [<ffffffff812c3df0>] ? shm_get_unmapped_area+0x20/0x20
[  163.984499]  [<ffffffff816caa14>] tracesys+0xdd/0xe2
Quoted 6 lines

This bug was introduced by commit af73e4d9506d3b797509f3c030e7dcd554f7d9c4 ("hugetlbfs: fix mmap failure in unaligned size request").

#4 lore raw
In reply to Li Zefan
Quoted 34 lines · Li Zefan
Cc Naoya Horiguchi <[email protected]>
Quoted 4 lines · Dave Jones

Just saw this on v3.9-11789-ge0fd9af while fuzz-testing.

[  163.917836] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[  163.918984] IP: [<ffffffff812c48ed>] newseg+0x10d/0x390

The patch below should fix it.

===========================

[PATCH] shm: fix null pointer deref when userspace specifies invalid hugepage size

Dave reported an oops triggered by trinity:

[  163.917836] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[  163.918984] IP: [<ffffffff812c48ed>] newseg+0x10d/0x390
[  163.919705] PGD cf8c1067 PUD cf8c2067 PMD 0
[  163.920326] Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC
[  163.929949] CPU: 2 PID: 7636 Comm: trinity-child2 Not tainted 3.9.0+#67
...
[  163.953629] Call Trace:
[  163.957706]  [<ffffffff812be322>] ipcget+0x182/0x380
[  163.962123]  [<ffffffff810b99a5>] ?trace_hardirqs_on_caller+0x115/0x1e0
[  163.966752]  [<ffffffff812c559a>] SyS_shmget+0x5a/0x60
[  163.971163]  [<ffffffff812c47e0>] ? shm_close+0x140/0x140
[  163.975590]  [<ffffffff812c3e60>] ? shm_release+0x50/0x50
[  163.979991]  [<ffffffff812c3df0>] ? shm_get_unmapped_area+0x20/0x20
[  163.984499]  [<ffffffff816caa14>] tracesys+0xdd/0xe2

This bug was introduced by commit af73e4d9506d3b797509f3c030e7dcd554f7d9c4 ("hugetlbfs: fix mmap failure in unaligned size request").

Quoted 6 lines

---

Previously it would return -ENODEV, but seems -EINVAL is more appropriate.

Should the user in fs/hugetlbfs/inode.c be updated as well? get_hstate_idx() is just returning -1 and it's only caller, hugetlb_file_setup(), returns -ENODEV.

Thanks,
Davidlohr