Linux-SCSI Archive on lore.kernel.org
 help / color / Atom feed
From: Denis Efremov <efremov@linux.com>
To: Joe Perches <joe@perches.com>,
	"Martin K. Petersen" <martin.petersen@oracle.com>,
	"James E.J. Bottomley" <jejb@linux.ibm.com>
Cc: linux-scsi@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] scsi: libcxgbi: use kvzalloc instead of opencoded kzalloc/vzalloc
Date: Sat, 1 Aug 2020 10:58:56 +0300
Message-ID: <47da22c0-d641-e94f-d38e-8af225ca115a@linux.com> (raw)
In-Reply-To: <c5a18804-3236-9688-2a3c-68184f0dd9e8@linux.com>



On 8/1/20 10:51 AM, Denis Efremov wrote:
> 
> 
> On 8/1/20 1:24 AM, Joe Perches wrote:
>> On Sat, 2020-08-01 at 01:10 +0300, Denis Efremov wrote:
>>>
>>> On 8/1/20 12:58 AM, Joe Perches wrote:
>>>> On Sat, 2020-08-01 at 00:55 +0300, Denis Efremov wrote:
>>>>> Remove cxgbi_alloc_big_mem(), cxgbi_free_big_mem() functions
>>>>> and use kvzalloc/kvfree instead.
>>>>
>>>> Sensible, thanks.
>>>>
>>>>> diff --git a/drivers/scsi/cxgbi/libcxgbi.c b/drivers/scsi/cxgbi/libcxgbi.c
>>>> []
>>>>> @@ -77,9 +77,9 @@ int cxgbi_device_portmap_create(struct cxgbi_device *cdev, unsigned int base,
>>>>>  {
>>>>>  	struct cxgbi_ports_map *pmap = &cdev->pmap;
>>>>>  
>>>>> -	pmap->port_csk = cxgbi_alloc_big_mem(max_conn *
>>>>> -					     sizeof(struct cxgbi_sock *),
>>>>> -					     GFP_KERNEL);
>>>>> +	pmap->port_csk = kvzalloc(array_size(max_conn,
>>>>> +					     sizeof(struct cxgbi_sock *)),
>>>>> +				  GFP_KERNEL);
>>>>
>>>> missing __GFP_NOWARN
>>>>
>>>
>>> kvmalloc_node adds __GFP_NOWARN internally to kmalloc call
>>> https://elixir.bootlin.com/linux/v5.8-rc4/source/mm/util.c#L568
>>
>> Only when there's a fallback, and the fallback does not.
>>
> 
> Sorry, Joe, I don't understand why do we need to add __GFP_NOWARN here.
> 
> We use GFP_KERNEL for allocation. cxgbi_alloc_big_mem adds __GFP_NOWARN only to kzalloc().
> kvzalloc is: kvmalloc(size, flags | __GFP_ZERO)
> kvmalloc is: kvmalloc_node(size, flags, NUMA_NO_NODE)
> kvmalloc_node: 
> 	if ((flags & GFP_KERNEL) != GFP_KERNEL)           // false, flags == GFP_KERNEL|__GFP_ZERO
> 		return kmalloc_node(size, flags, node);
> 
> 	if (size > PAGE_SIZE) {
> 		kmalloc_flags |= __GFP_NOWARN;            // add __GFP_NOWARN for "big" allocations
> 
> 		if (!(kmalloc_flags & __GFP_RETRY_MAYFAIL))
> 			kmalloc_flags |= __GFP_NORETRY;
> 	}
> 
> 	// kmalloc_flags == GFP_KERNEL|__GFP_ZERO|__GFP_NOWARN if size > PAGE_SIZE
> 	ret = kmalloc_node(size, kmalloc_flags, node);    
> 
> 	if (ret || size <= PAGE_SIZE)
> 		return ret;
> 
> 	// flags == GFP_KERNEL|__GFP_ZERO
> 	return __vmalloc_node(size, 1, flags, node,
> 			__builtin_return_address(0));
> 
> So, to my understanding the difference is only that cxgbi_alloc_big_mem adds __GFP_NOWARN
> to kzalloc unconditionally, kvzalloc adds __GFP_NOWARN to kmalloc_node if size > PAGE_SIZE.
> We use: max_conn * sizeof(struct cxgbi_sock *) in allocation. max_conn can be either
> CXGBI_MAX_CONN or CXGB4I_MAX_CONN. CXGBI_MAX_CONN == 16384, CXGB4I_MAX_CONN == 16384.
> Thus the allocation is bigger than PAGE_SIZE and kvmalloc_node adds __GFP_NOWARN to the
> kmalloc_node call. Maybe I missed something?
> 

Of course max_conn can be smaller than CXGBI_MAX_CONN and CXGB4I_MAX_CONN and size < PAGE_SIZE.
However, it's by design not to warn on "small" allocations in kvmalloc_node and it will not even
try to call __vmalloc_node if "small" allocation will fail with kmalloc_node. I think that this
behavior fits the cxgbi_device_portmap_create() context call.

  reply index

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-31 21:55 Denis Efremov
2020-07-31 21:58 ` Joe Perches
2020-07-31 22:10   ` Denis Efremov
2020-07-31 22:24     ` Joe Perches
2020-08-01  7:51       ` Denis Efremov
2020-08-01  7:58         ` Denis Efremov [this message]
2020-08-01  8:10         ` Joe Perches
2020-08-01 13:28           ` Denis Efremov
2020-08-01 13:31 ` [PATCH v2] " Denis Efremov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=47da22c0-d641-e94f-d38e-8af225ca115a@linux.com \
    --to=efremov@linux.com \
    --cc=jejb@linux.ibm.com \
    --cc=joe@perches.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-scsi@vger.kernel.org \
    --cc=martin.petersen@oracle.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-SCSI Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-scsi/0 linux-scsi/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-scsi linux-scsi/ https://lore.kernel.org/linux-scsi \
		linux-scsi@vger.kernel.org
	public-inbox-index linux-scsi

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-scsi


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git