Re: fix iounmap and a pageattr memleak (x86 and x86-64)

[Dave Hansen]

Andrea Arcangeli wrote:
> > I'll see what I can do to get some backtraces of the __pg_prot(0) &&
> > page->mapped cases.

These appear to be the 2 most common paths:  cache_init_objs() and 
kfree().  But, the path that I'm hitting now appears to be that 
something got page_count(kpte_page) to -1 (*NOT* page->_count), and then 
the BUG_ON() trips after a get_page() is done on it.  I'm tracking this 
down now.

kpte: c0011000
address: c1a00000
kpte_page: c1000264
split: ffffffff
pgprot_val(prot): 00000000
pgprot_val(PAGE_KERNEL): 00000163
(pte_val(*kpte) & _PAGE_PSE): 00000000
path: 1
Badness in __change_page_attr at arch/i386/mm/pageattr.c:156
 [<c0114e43>] __change_page_attr+0x443/0x5a4
 [<c01150d9>] kernel_map_pages+0x31/0x80
 [<c0114fda>] change_page_attr+0x36/0x54
 [<c01150f3>] kernel_map_pages+0x4b/0x80
 [<c013de18>] cache_init_objs+0x194/0x1d8
 [<c013e014>] cache_grow+0xe8/0x154
 [<c013e74a>] cache_alloc_refill+0x226/0x274
 [<c013edb3>] __kmalloc+0x8b/0xbc
 [<c0181302>] proc_create+0x76/0xcc
 [<c01814c3>] create_proc_entry+0x6b/0xb4
 [<c0121c3b>] register_proc_table+0xcb/0x110
 [<c0121c67>] register_proc_table+0xf7/0x110
 [<c0420994>] sysctl_init+0x10/0x1c
 [<c0413934>] do_basic_setup+0x14/0x20
 [<c01004ec>] init+0xa8/0x15c
 [<c0100444>] init+0x0/0x15c
 [<c0102305>] kernel_thread_helper+0x5/0xc

kpte: c0006978
address: c052f000
kpte_page: c10000d8
split: ffffffff
pgprot_val(prot): 00000000
pgprot_val(PAGE_KERNEL): 00000163
(pte_val(*kpte) & _PAGE_PSE): 00000000
path: 1
Badness in __change_page_attr at arch/i386/mm/pageattr.c:156
 [<c0114e43>] __change_page_attr+0x443/0x5a4
 [<c02367db>] scsi_run_queue+0xaf/0xb8
 [<c0114fda>] change_page_attr+0x36/0x54
 [<c01150f3>] kernel_map_pages+0x4b/0x80
 [<c013e422>] cache_free_debugcheck+0x2a6/0x2c0
 [<c02385a5>] scsi_probe_and_add_lun+0x135/0x188
 [<c013f00b>] kfree+0x9f/0xdc
 [<c02385a5>] scsi_probe_and_add_lun+0x135/0x188
 [<c02385a5>] scsi_probe_and_add_lun+0x135/0x188
 [<c0238ac7>] scsi_scan_target+0x63/0xbc
 [<c0238b60>] scsi_scan_channel+0x40/0x68
 [<c0238bfe>] scsi_scan_host_selected+0x76/0xb0
 [<c0238c4a>] scsi_scan_host+0x12/0x18
 [<c024b7ff>] ahc_linux_register_host+0x283/0x290
 [<c01857e9>] sysfs_create_file+0x41/0x48
 [<c01e4392>] pci_create_newid_file+0x1a/0x24
 [<c01e470a>] pci_populate_driver_dir+0xa/0x10
 [<c01e4786>] pci_register_driver+0x76/0x88
 [<c024a867>] ahc_linux_detect+0x3b/0x60
 [<c042750a>] ahc_linux_init+0xa/0x30
 [<c04138ca>] do_initcalls+0x66/0xbc
 [<c041393e>] do_basic_setup+0x1e/0x20
 [<c01004ec>] init+0xa8/0x15c
 [<c0100444>] init+0x0/0x15c
 [<c0102305>] kernel_thread_helper+0x5/0xc
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/