2019-11-11 22:51:49

by Ralph Campbell

Subject: [PATCH] mm/debug:

When dumping struct page information, __dump_page() prints the page type
with a trailing blank followed by the page flags on a separate line:

anon
flags: 0x100000000090034(uptodate|lru|active|head|swapbacked)

Fix this by using pr_cont() instead of pr_warn() to get a single line:

anon flags: 0x100000000090034(uptodate|lru|active|head|swapbacked)

Signed-off-by: Ralph Campbell <[email protected]>
---
mm/debug.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/debug.c b/mm/debug.c
index 8345bb6e4769..752c78721ea0 100644
--- a/mm/debug.c
+++ b/mm/debug.c
@@ -87,7 +87,7 @@ void __dump_page(struct page *page, const char *reason)
}
BUILD_BUG_ON(ARRAY_SIZE(pageflag_names) != __NR_PAGEFLAGS + 1);

- pr_warn("flags: %#lx(%pGp)\n", page->flags, &page->flags);
+ pr_cont("flags: %#lx(%pGp)\n", page->flags, &page->flags);

hex_only:
print_hex_dump(KERN_WARNING, "raw: ", DUMP_PREFIX_NONE, 32,
--
2.20.1

2019-11-11 23:03:52

by Andrew Morton

Subject: Re: [PATCH] mm/debug:

On Mon, 11 Nov 2019 14:49:35 -0800 Ralph Campbell <[email protected]> wrote:

> When dumping struct page information, __dump_page() prints the page type
> with a trailing blank followed by the page flags on a separate line:
>
> anon
> flags: 0x100000000090034(uptodate|lru|active|head|swapbacked)
>
> Fix this by using pr_cont() instead of pr_warn() to get a single line:
>
> anon flags: 0x100000000090034(uptodate|lru|active|head|swapbacked)
>
> Signed-off-by: Ralph Campbell <[email protected]>
> ---
> mm/debug.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/debug.c b/mm/debug.c
> index 8345bb6e4769..752c78721ea0 100644
> --- a/mm/debug.c
> +++ b/mm/debug.c
> @@ -87,7 +87,7 @@ void __dump_page(struct page *page, const char *reason)
> }
> BUILD_BUG_ON(ARRAY_SIZE(pageflag_names) != __NR_PAGEFLAGS + 1);
>
> - pr_warn("flags: %#lx(%pGp)\n", page->flags, &page->flags);
> + pr_cont("flags: %#lx(%pGp)\n", page->flags, &page->flags);
>
> hex_only:
> print_hex_dump(KERN_WARNING, "raw: ", DUMP_PREFIX_NONE, 32,

This is the case if PageAnon || PageKsm || mapping. If it is, say,
PageSlab then we effectively do

pr_warn("stuff-with-no-newline");
pr_cont("\n");
pr_cont("flags: ...\n");

does this work OK? what facility level will that "flags: " line get?

2019-11-11 23:44:44

by Ralph Campbell

Subject: Re: [PATCH] mm/debug:

On 11/11/19 3:02 PM, Andrew Morton wrote:
> On Mon, 11 Nov 2019 14:49:35 -0800 Ralph Campbell <[email protected]> wrote:
>
>> When dumping struct page information, __dump_page() prints the page type
>> with a trailing blank followed by the page flags on a separate line:
>>
>> anon
>> flags: 0x100000000090034(uptodate|lru|active|head|swapbacked)
>>
>> Fix this by using pr_cont() instead of pr_warn() to get a single line:
>>
>> anon flags: 0x100000000090034(uptodate|lru|active|head|swapbacked)
>>
>> Signed-off-by: Ralph Campbell <[email protected]>
>> ---
>> mm/debug.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/mm/debug.c b/mm/debug.c
>> index 8345bb6e4769..752c78721ea0 100644
>> --- a/mm/debug.c
>> +++ b/mm/debug.c
>> @@ -87,7 +87,7 @@ void __dump_page(struct page *page, const char *reason)
>> }
>> BUILD_BUG_ON(ARRAY_SIZE(pageflag_names) != __NR_PAGEFLAGS + 1);
>>
>> - pr_warn("flags: %#lx(%pGp)\n", page->flags, &page->flags);
>> + pr_cont("flags: %#lx(%pGp)\n", page->flags, &page->flags);
>>
>> hex_only:
>> print_hex_dump(KERN_WARNING, "raw: ", DUMP_PREFIX_NONE, 32,
>
> This is the case if PageAnon || PageKsm || mapping. If it is, say,
> PageSlab then we effectively do
>
> pr_warn("stuff-with-no-newline");
> pr_cont("\n");
> pr_cont("flags: ...\n");
>
> does this work OK? what facility level will that "flags: " line get?

I don't see a "\n" in the "mapping" case but the
if (mapping->host && mapping->host->i_dentry.first)
pr_warn("name:\"%pd\" ", dentry)
would end up on a new line.
Ugh. I guess the dentry name could be fairly long.
I guess I will just convert to using "\n".

AFAIK, there is no locking between pr_warn() and pr_cont() so the latter
could get appended to any facility level. But that isn't a problem I
plan to fix.