| Message ID | 20190618124352.28307-1-colin.king@canonical.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | mm: idle-page: fix oops because end_pfn is larger than max_pfn | expand |
On Tue, 18 Jun 2019 13:43:52 +0100 Colin King <colin.king@canonical.com> wrote: > From: Colin Ian King <colin.king@canonical.com> > > Currently the calcuation of end_pfn can round up the pfn number to > more than the actual maximum number of pfns, causing an Oops. Fix > this by ensuring end_pfn is never more than max_pfn. > > This can be easily triggered when on systems where the end_pfn gets > rounded up to more than max_pfn using the idle-page stress-ng > stress test: > cc Vladimir. This seems rather obvious - I'm wondering if the code was that way for some subtle reason? (I'll add a cc:stable to this) From: Colin Ian King <colin.king@canonical.com> Subject: mm/page_idle.c: fix oops because end_pfn is larger than max_pfn Currently the calcuation of end_pfn can round up the pfn number to more than the actual maximum number of pfns, causing an Oops. Fix this by ensuring end_pfn is never more than max_pfn. This can be easily triggered when on systems where the end_pfn gets rounded up to more than max_pfn using the idle-page stress-ng stress test: sudo stress-ng --idle-page 0 [ 3812.222790] BUG: unable to handle kernel paging request at 00000000000020d8 [ 3812.224341] #PF error: [normal kernel read fault] [ 3812.225144] PGD 0 P4D 0 [ 3812.225626] Oops: 0000 [#1] SMP PTI [ 3812.226264] CPU: 1 PID: 11039 Comm: stress-ng-idle- Not tainted 5.0.0-5-generic #6-Ubuntu [ 3812.227643] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1ubuntu1 04/01/2014 [ 3812.229286] RIP: 0010:page_idle_get_page+0xc8/0x1a0 [ 3812.230173] Code: 0f b1 0a 75 7d 48 8b 03 48 89 c2 48 c1 e8 33 83 e0 07 48 c1 ea 36 48 8d 0c 40 4c 8d 24 88 49 c1 e4 07 4c 03 24 d5 00 89 c3 be <49> 8b 44 24 58 48 8d b8 80 a1 02 00 e8 07 d5 77 00 48 8b 53 08 48 [ 3812.234641] RSP: 0018:ffffafd7c672fde8 EFLAGS: 00010202 [ 3812.235792] RAX: 0000000000000005 RBX: ffffe36341fff700 RCX: 000000000000000f [ 3812.237739] RDX: 0000000000000284 RSI: 0000000000000275 RDI: 0000000001fff700 [ 3812.239225] RBP: ffffafd7c672fe00 R08: ffffa0bc34056410 R09: 0000000000000276 [ 3812.241027] R10: ffffa0bc754e9b40 R11: ffffa0bc330f6400 R12: 0000000000002080 [ 3812.242555] R13: ffffe36341fff700 R14: 0000000000080000 R15: ffffa0bc330f6400 [ 3812.244073] FS: 00007f0ec1ea5740(0000) GS:ffffa0bc7db00000(0000) knlGS:0000000000000000 [ 3812.245968] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 3812.247162] CR2: 00000000000020d8 CR3: 0000000077d68000 CR4: 00000000000006e0 [ 3812.249045] Call Trace: [ 3812.249625] page_idle_bitmap_write+0x8c/0x140 [ 3812.250567] sysfs_kf_bin_write+0x5c/0x70 [ 3812.251406] kernfs_fop_write+0x12e/0x1b0 [ 3812.252282] __vfs_write+0x1b/0x40 [ 3812.253002] vfs_write+0xab/0x1b0 [ 3812.253941] ksys_write+0x55/0xc0 [ 3812.254660] __x64_sys_write+0x1a/0x20 [ 3812.255446] do_syscall_64+0x5a/0x110 [ 3812.256254] entry_SYSCALL_64_after_hwframe+0x44/0xa9 --- a/mm/page_idle.c~mm-idle-page-fix-oops-because-end_pfn-is-larger-than-max_pfn +++ a/mm/page_idle.c @@ -136,7 +136,7 @@ static ssize_t page_idle_bitmap_read(str end_pfn = pfn + count * BITS_PER_BYTE; if (end_pfn > max_pfn) - end_pfn = ALIGN(max_pfn, BITMAP_CHUNK_BITS); + end_pfn = max_pfn; for (; pfn < end_pfn; pfn++) { bit = pfn % BITMAP_CHUNK_BITS; @@ -181,7 +181,7 @@ static ssize_t page_idle_bitmap_write(st end_pfn = pfn + count * BITS_PER_BYTE; if (end_pfn > max_pfn) - end_pfn = ALIGN(max_pfn, BITMAP_CHUNK_BITS); + end_pfn = max_pfn; for (; pfn < end_pfn; pfn++) { bit = pfn % BITMAP_CHUNK_BITS;
On Tue, Jun 18, 2019 at 12:45:02PM -0700, Andrew Morton wrote: > On Tue, 18 Jun 2019 13:43:52 +0100 Colin King <colin.king@canonical.com> wrote: > > > From: Colin Ian King <colin.king@canonical.com> > > > > Currently the calcuation of end_pfn can round up the pfn number to > > more than the actual maximum number of pfns, causing an Oops. Fix > > this by ensuring end_pfn is never more than max_pfn. > > > > This can be easily triggered when on systems where the end_pfn gets > > rounded up to more than max_pfn using the idle-page stress-ng > > stress test: > > > > cc Vladimir. This seems rather obvious - I'm wondering if the code was > that way for some subtle reason? No subtle reason at all - just a bug. The patch looks good to me, Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
diff --git a/mm/page_idle.c b/mm/page_idle.c index 0b39ec0c945c..295512465065 100644 --- a/mm/page_idle.c +++ b/mm/page_idle.c @@ -136,7 +136,7 @@ static ssize_t page_idle_bitmap_read(struct file *file, struct kobject *kobj, end_pfn = pfn + count * BITS_PER_BYTE; if (end_pfn > max_pfn) - end_pfn = ALIGN(max_pfn, BITMAP_CHUNK_BITS); + end_pfn = max_pfn; for (; pfn < end_pfn; pfn++) { bit = pfn % BITMAP_CHUNK_BITS; @@ -181,7 +181,7 @@ static ssize_t page_idle_bitmap_write(struct file *file, struct kobject *kobj, end_pfn = pfn + count * BITS_PER_BYTE; if (end_pfn > max_pfn) - end_pfn = ALIGN(max_pfn, BITMAP_CHUNK_BITS); + end_pfn = max_pfn; for (; pfn < end_pfn; pfn++) { bit = pfn % BITMAP_CHUNK_BITS;