Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   "Jonathan Lemon" <jonathan.lemon@gmail.com>
To:     "Ivan Khoronzhuk" <ivan.khoronzhuk@linaro.org>
Cc:     magnus.karlsson@intel.com, bjorn.topel@intel.com,
        davem@davemloft.net, hawk@kernel.org, john.fastabend@gmail.com,
        jakub.kicinski@netronome.com, daniel@iogearbox.net,
        netdev@vger.kernel.org, bpf@vger.kernel.org,
        xdp-newbies@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH bpf-next 2/3] xdp: xdp_umem: replace kmap on vmap for umem
 map
Date:   Tue, 13 Aug 2019 11:33:18 -0700
Message-ID: <BC98A490-2892-452B-AC3E-C9B9F9BA121C@gmail.com>
In-Reply-To: <20190813183023.GA2856@khorivan>
References: <20190813102318.5521-1-ivan.khoronzhuk@linaro.org>
 <20190813102318.5521-3-ivan.khoronzhuk@linaro.org>
 <9F98648A-8654-4767-97B5-CF4BC939393C@flugsvamp.com>
 <20190813183023.GA2856@khorivan>
MIME-Version: 1.0
Content-Type: text/plain; format=flowed
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 13 Aug 2019, at 11:30, Ivan Khoronzhuk wrote:

> On Tue, Aug 13, 2019 at 10:42:18AM -0700, Jonathan Lemon wrote:
>>
>>
>> On 13 Aug 2019, at 3:23, Ivan Khoronzhuk wrote:
>>
>>> For 64-bit there is no reason to use vmap/vunmap, so use 
>>> page_address
>>> as it was initially. For 32 bits, in some apps, like in samples
>>> xdpsock_user.c when number of pgs in use is quite big, the kmap
>>> memory can be not enough, despite on this, kmap looks like is
>>> deprecated in such cases as it can block and should be used rather
>>> for dynamic mm.
>>>
>>> Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
>>
>> Seems a bit overkill - if not high memory, kmap() falls back
>> to just page_address(), unlike vmap().
>
>> -- Jonathan
>
> So, as kmap has limitation... if I correctly understood, you propose
> to avoid macros and do smth like kmap:
>
> 	void *addr;
> 	if (!PageHighMem(&umem->pgs[i]))
> 		addr =  page_address(page);
> 	else
> 		addr = vmap(&umem->pgs[i], 1, VM_MAP, PAGE_KERNEL);
>
> 	umem->pages[i].addr = addr;
>
> and while unmap
>
> 	if (!PageHighMem(&umem->pgs[i]))
> 		vunmap(umem->pages[i].addr);
>
> I can try it, and add this in v2 if no objection.

Seems like a reasonable compromise to me.
-- 
Jonathan


>
>>
>>> ---
>>> net/xdp/xdp_umem.c | 16 ++++++++++++----
>>> 1 file changed, 12 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/net/xdp/xdp_umem.c b/net/xdp/xdp_umem.c
>>> index a0607969f8c0..907c9019fe21 100644
>>> --- a/net/xdp/xdp_umem.c
>>> +++ b/net/xdp/xdp_umem.c
>>> @@ -14,7 +14,7 @@
>>> #include <linux/netdevice.h>
>>> #include <linux/rtnetlink.h>
>>> #include <linux/idr.h>
>>> -#include <linux/highmem.h>
>>> +#include <linux/vmalloc.h>
>>>
>>> #include "xdp_umem.h"
>>> #include "xsk_queue.h"
>>> @@ -167,10 +167,12 @@ void xdp_umem_clear_dev(struct xdp_umem *umem)
>>>
>>> static void xdp_umem_unmap_pages(struct xdp_umem *umem)
>>> {
>>> +#if BITS_PER_LONG == 32
>>> 	unsigned int i;
>>>
>>> 	for (i = 0; i < umem->npgs; i++)
>>> -		kunmap(umem->pgs[i]);
>>> +		vunmap(umem->pages[i].addr);
>>> +#endif
>>> }
>>>
>>> static void xdp_umem_unpin_pages(struct xdp_umem *umem)
>>> @@ -378,8 +380,14 @@ static int xdp_umem_reg(struct xdp_umem *umem, 
>>> struct xdp_umem_reg *mr)
>>> 		goto out_account;
>>> 	}
>>>
>>> -	for (i = 0; i < umem->npgs; i++)
>>> -		umem->pages[i].addr = kmap(umem->pgs[i]);
>>> +	for (i = 0; i < umem->npgs; i++) {
>>> +#if BITS_PER_LONG == 32
>>> +		umem->pages[i].addr = vmap(&umem->pgs[i], 1, VM_MAP,
>>> +					   PAGE_KERNEL);
>>> +#else
>>> +		umem->pages[i].addr = page_address(umem->pgs[i]);
>>> +#endif
>>> +	}
>>>
>>> 	return 0;
>>>
>>> -- 
>>> 2.17.1
>
> -- 
> Regards,
> Ivan Khoronzhuk