CMA, isolate: get warning in page_isolation.c:235 test_pages_isolated

DISCUSSION 朱辉 2 participants 2 messages Nov 28, 2014 Memory management

Share & tools

朱辉

author

2014-11-28 03:15 UTC #0 lore raw

SGkgZ3V5cywKCkFmdGVyIEkgYmFjayBwb3J0aW5nIHlvdXIgcGF0Y2hlczoKbW0vcGFnZV9hbGxv YzogZml4IGluY29ycmVjdCBpc29sYXRpb24gYmVoYXZpb3IgYnkgcmVjaGVja2luZyBtaWdyYXRl dHlwZQptbS9wYWdlX2FsbG9jOiBhZGQgZnJlZXBhZ2Ugb24gaXNvbGF0ZSBwYWdlYmxvY2sgdG8g Y29ycmVjdCBidWRkeSBsaXN0Cm1tL3BhZ2VfYWxsb2M6IG1vdmUgZnJlZXBhZ2UgY291bnRpbmcg bG9naWMgdG8gX19mcmVlX29uZV9wYWdlKCkKbW0vcGFnZV9hbGxvYzogcmVzdHJpY3QgbWF4IG9y ZGVyIG9mIG1lcmdpbmcgb24gaXNvbGF0ZWQgcGFnZWJsb2NrCm1tOiBwYWdlX2FsbG9jOiBzdG9y ZSB1cGRhdGVkIHBhZ2UgbWlncmF0ZXR5cGUgdG8gYXZvaWQgbWlzdXNpbmcgc3RhbGUgdmFsdWUK bW06IHBhZ2VfaXNvbGF0aW9uOiBjaGVjayBwZm4gdmFsaWRpdHkgYmVmb3JlIGFjY2Vzcwp0byAz LjEwIGxpbnV4IGtlcm5lbC4KSSBhbHNvIHVzZSB0aGUgQ01BX0FHR1JFU1NJVkUgcGF0Y2hlcyBp biBodHRwczovL2xrbWwub3JnL2xrbWwvMjAxNC8xMC8xNS82MjMuCgpJIGdvdDoKWzY4MTIxLjc3 MDY5OUAyXSAtLS0tLS0tLS0tLS1bIGN1dCBoZXJlIF0tLS0tLS0tLS0tLS0KWzY4MTIxLjc3NDU5 MkAyXSBXQVJOSU5HOiBhdCAvaG9tZS90ZWF3YXRlci9jb21tb24vbW0vcGFnZV9pc29sYXRpb24u YzoyMzUgdGVzdF9wYWdlc19pc29sYXRlZCsweDEwOC8weDIwOCgpCls2ODEyMS43OTM5MTFAMl0g Q1BVOiAyIFBJRDogMjcxMSBDb21tOiBrdGhyZWFkX3h4eCBUYWludGVkOiBQICAgICAgICAgICBP IDMuMTAuMzMtMjUwNjQ0LWdjZmQ5M2Y4LWRpcnR5ICMxODQKWzY4MTIxLjgwMzYzMkAyXSBbPGMw MDE2ZGU0Pl0gKHVud2luZF9iYWNrdHJhY2UrMHgwLzB4MTI4KSBmcm9tIFs8YzAwMTMzNjA+XSAo c2hvd19zdGFjaysweDIwLzB4MjQpCls2ODEyMS44MTIzNzlAMl0gWzxjMDAxMzM2MD5dIChzaG93 X3N0YWNrKzB4MjAvMHgyNCkgZnJvbSBbPGMwNzQ1NTNjPl0gKGR1bXBfc3RhY2srMHgyMC8weDI4 KQpbNjgxMjEuODIwNjEyQDJdIFs8YzA3NDU1M2M+XSAoZHVtcF9zdGFjaysweDIwLzB4MjgpIGZy b20gWzxjMDAyZjJiOD5dICh3YXJuX3Nsb3dwYXRoX2NvbW1vbisweDVjLzB4N2MpCls2ODEyMS44 Mjk3MTJAMl0gWzxjMDAyZjJiOD5dICh3YXJuX3Nsb3dwYXRoX2NvbW1vbisweDVjLzB4N2MpIGZy b20gWzxjMDAyZjMwND5dICh3YXJuX3Nsb3dwYXRoX251bGwrMHgyYy8weDM0KQpbNjgxMjEuODM5 NTA4QDJdIFs8YzAwMmYzMDQ+XSAod2Fybl9zbG93cGF0aF9udWxsKzB4MmMvMHgzNCkgZnJvbSBb PGMwMTFmMzI0Pl0gKHRlc3RfcGFnZXNfaXNvbGF0ZWQrMHgxMDgvMHgyMDgpCls2ODEyMS44NDkz OTNAMl0gWzxjMDExZjMyND5dICh0ZXN0X3BhZ2VzX2lzb2xhdGVkKzB4MTA4LzB4MjA4KSBmcm9t IFs8YzAwZTI0ZDg+XSAoYWxsb2NfY29udGlnX3JhbmdlKzB4MjA4LzB4MmIwKQpbNjgxMjEuODU5 NDQ3QDJdIFs8YzAwZTI0ZDg+XSAoYWxsb2NfY29udGlnX3JhbmdlKzB4MjA4LzB4MmIwKSBmcm9t IFs8YzAzMjBkNDQ+XSAoZG1hX2FsbG9jX2Zyb21fY29udGlndW91cysweDE1Yy8weDI0YykKCkxv b2tzIGl0IGhhcyBzb21lIHJhY2UgaXNzdWUgYmV0d2VlbiBwYWdlIGlzb2xhdGlvbiBhbmQgZnJl ZSBwYXRoIGFmdGVyIHRoZXNlIHBhdGNoZXMuCkFuZCBJIGNoZWNrZWQgdGhlIGZyZWUgcGF0aCBi dXQgZm91bmQgbm90aGluZy4KCkkgd29ycmllZCB0aGF0IGl0IHN0aWxsIGhhcyBzb21lIHJhY2Ug aXNzdWUgYmV0d2VlbiBwYWdlIGlzb2xhdGlvbiBhbmQgc29tZXRoaW5nIGluIHVwc3RyZWFtLiAg T3IgSSBtaXNzZWQgc29tZSBwYXRjaGVzPwpJZiB3ZSBjYW5ub3QgaGFuZGxlIHRoaXMgaXNzdWUg aW4gYSBzaG9ydCB0aW1lLCBJIHN1Z2dlc3QgYWRkIHRoZSAibW92ZV9mcmVlcGFnZXMiIGNvZGUg YmFjayB0byBfX3Rlc3RfcGFnZV9pc29sYXRlZF9pbl9wYWdlYmxvY2suCgpUaGFua3MsCkh1aQoK IA==

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Weijie Yang

2014-11-28 04:48 UTC #1 lore raw

Quoted 9 lines

-----Original Message-----
From: 朱辉 [mailto:[email protected]]
Sent: Friday, November 28, 2014 11:15 AM
To: [email protected]; [email protected]
Cc: Hui Zhu; [email protected]
Subject: CMA, isolate: get warning in page_isolation.c:235 test_pages_isolated
Hi guys,

Quoted 7 lines

After I back porting your patches:
mm/page_alloc: fix incorrect isolation behavior by rechecking migratetype mm/page_alloc: add freepage on isolate pageblock to correct buddy list mm/page_alloc: move freepage counting logic to __free_one_page() mm/page_alloc: restrict max order of merging on isolated pageblock
mm: page_alloc: store updated page migratetype to avoid misusing stale value
mm: page_isolation: check pfn validity before access

My patch is not proper, it doesn't consider the steal freepages in alloc path. So it should be dropped(I will send a notice email to akpm).

I am now working on its v2 patch and will resend it soon.

Thanks

Quoted 15 lines

to 3.10 linux kernel.
I also use the CMA_AGGRESSIVE patches in https://lkml.org/lkml/2014/10/15/623.
I got:
[68121.770699@2] ------------[ cut here ]------------
[68121.774592@2] WARNING: at /home/teawater/common/mm/page_isolation.c:235 test_pages_isolated+0x108/0x208()
[68121.793911@2] CPU: 2 PID: 2711 Comm: kthread_xxx Tainted: P O 3.10.33-250644-gcfd93f8-dirty #184
[68121.803632@2] [<c0016de4>] (unwind_backtrace+0x0/0x128) from [<c0013360>] (show_stack+0x20/0x24) [68121.812379@2] [<c0013360>] (show_stack+0x20/0x24) from [<c074553c>] (dump_stack+0x20/0x28) [68121.820612@2] [<c074553c>] (dump_stack+0x20/0x28) from [<c002f2b8>] (warn_slowpath_common+0x5c/0x7c) [68121.829712@2] [<c002f2b8>] (warn_slowpath_common+0x5c/0x7c) from [<c002f304>] (warn_slowpath_null+0x2c/0x34) [68121.839508@2] [<c002f304>] (warn_slowpath_null+0x2c/0x34) from [<c011f324>] (test_pages_isolated+0x108/0x208) [68121.849393@2] [<c011f324>] (test_pages_isolated+0x108/0x208) from [<c00e24d8>] (alloc_contig_range+0x208/0x2b0) [68121.859447@2] [<c00e24d8>] (alloc_contig_range+0x208/0x2b0) from [<c0320d44>] (dma_alloc_from_contiguous+0x15c/0x24c)

Quoted 10 lines

Looks it has some race issue between page isolation and free path after these patches. And I checked the free path but found nothing.
I worried that it still has some race issue between page isolation and something in upstream. Or I missed some patches? If we cannot handle this issue in a short time, I suggest add the "move_freepages" code back to __test_page_isolated_in_pageblock.
Thanks,
Hui