lkml.org 
[lkml]   [2019]   [Mar]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] kmemleak: skip scanning holes in the .bss section
On Tue, 12 Mar 2019 15:14:12 -0400 Qian Cai <cai@lca.pw> wrote:

> The commit 2d4f567103ff ("KVM: PPC: Introduce kvm_tmp framework") adds
> kvm_tmp[] into the .bss section and then free the rest of unused spaces
> back to the page allocator.
>
> kernel_init
> kvm_guest_init
> kvm_free_tmp
> free_reserved_area
> free_unref_page
> free_unref_page_prepare
>
> With DEBUG_PAGEALLOC=y, it will unmap those pages from kernel. As the
> result, kmemleak scan will trigger a panic below when it scans the .bss
> section with unmapped pages.
>
> Since this is done way before the first kmemleak_scan(), just go
> lockless to make the implementation simple and skip those pages when
> scanning the .bss section. Later, those pages could be tracked by
> kmemleak again once allocated by the page allocator. Overall, this is
> such a special case, so no need to make it a generic to let kmemleak
> gain an ability to skip blocks in scan_large_block().
>
> BUG: Unable to handle kernel data access at 0xc000000001610000
> Faulting instruction address: 0xc0000000003cc178
> Oops: Kernel access of bad area, sig: 11 [#1]
> LE PAGE_SIZE=64K MMU=Hash SMP NR_CPUS=256 DEBUG_PAGEALLOC NUMA pSeries
> CPU: 3 PID: 130 Comm: kmemleak Kdump: loaded Not tainted 5.0.0+ #9
> REGS: c0000004b05bf940 TRAP: 0300 Not tainted (5.0.0+)
> NIP [c0000000003cc178] scan_block+0xa8/0x190
> LR [c0000000003cc170] scan_block+0xa0/0x190
> Call Trace:
> [c0000004b05bfbd0] [c0000000003cc170] scan_block+0xa0/0x190 (unreliable)
> [c0000004b05bfc30] [c0000000003cc2c0] scan_large_block+0x60/0xa0
> [c0000004b05bfc70] [c0000000003ccc64] kmemleak_scan+0x254/0x960
> [c0000004b05bfd40] [c0000000003cdd50] kmemleak_scan_thread+0xec/0x12c
> [c0000004b05bfdb0] [c000000000104388] kthread+0x1b8/0x1c0
> [c0000004b05bfe20] [c00000000000b364] ret_from_kernel_thread+0x5c/0x78
> Instruction dump:
> 7fa3eb78 4844667d 60000000 60000000 60000000 60000000 3bff0008 7fbcf840
> 409d00b8 4bfffeed 2fa30000 409e00ac <e87f0000> e93e0128 7fa91840
> 419dffdc
>

hm, yes, this is super crude. I guess we can turn it into something
more sophisticated if another caller is identified.

> --- a/mm/kmemleak.c
> +++ b/mm/kmemleak.c
> @@ -237,6 +237,10 @@ static int kmemleak_skip_disable;
> /* If there are leaks that can be reported */
> static bool kmemleak_found_leaks;
>
> +/* Skip scanning of a range in the .bss section. */
> +static void *bss_hole_start;
> +static void *bss_hole_stop;
> +
> static bool kmemleak_verbose;
> module_param_named(verbose, kmemleak_verbose, bool, 0600);
>
> @@ -1265,6 +1269,18 @@ void __ref kmemleak_ignore_phys(phys_addr_t phys)
> }
> EXPORT_SYMBOL(kmemleak_ignore_phys);
>
> +/**
> + * kmemleak_bss_hole - skip scanning a range in the .bss section
> + *
> + * @start: start of the range
> + * @stop: end of the range
> + */
> +void kmemleak_bss_hole(void *start, void *stop)
> +{
> + bss_hole_start = start;
> + bss_hole_stop = stop;
> +}

I'll make this __init.

> /*
> * Update an object's checksum and return true if it was modified.
> */
> @@ -1531,7 +1547,14 @@ static void kmemleak_scan(void)
>
> /* data/bss scanning */
> scan_large_block(_sdata, _edata);
> - scan_large_block(__bss_start, __bss_stop);
> +
> + if (bss_hole_start) {
> + scan_large_block(__bss_start, bss_hole_start);
> + scan_large_block(bss_hole_stop, __bss_stop);
> + } else {
> + scan_large_block(__bss_start, __bss_stop);
> + }
> +
> scan_large_block(__start_ro_after_init, __end_ro_after_init);
>
> #ifdef CONFIG_SMP

\
 
 \ /
  Last update: 2019-03-12 22:10    [W:0.124 / U:0.112 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site