Re: [Em28xx] ehci_hcd map_single: unable to map unsafe buffer on a standard NSLU2

Markus Rechberger a écrit :
> On 9/13/07, David & Anne-Sophie <[email protected]> wrote:
>   
>> Markus Rechberger a écrit :
>>     
>>> On 9/13/07, David & Anne-Sophie <[email protected]> wrote:
>>>
>>>       
>>>> Markus Rechberger a écrit :
>>>>
>>>>         
>>>>> On 9/12/07, David & Anne-Sophie <[email protected]> wrote:
>>>>>
>>>>>
>>>>>           
>>>>>> Markus Rechberger a écrit :
>>>>>>
>>>>>>
>>>>>>             
>>>>>>> On 9/11/07, David & Anne-Sophie <[email protected]> wrote:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>               
>>>>>>>> Markus Rechberger a écrit :
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>                 
>>>>>>>>> On 9/10/07, David & Anne-Sophie <[email protected]> wrote:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>                   
>>>>>>>>>> David & Anne-Sophie a écrit :
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>                     
>>>>>>>>>>> Hello all,
>>>>>>>>>>> I am trying to use my hybrid USB tuner on my NSLU2.
>>>>>>>>>>> Tuner=Terratec cinergy XS
>>>>>>>>>>> Driver=http://mcentral.de/wiki/index.php/Em2880
>>>>>>>>>>> I plugged it directly to the DISK-2 connector. On DISK-1 there is
>>>>>>>>>>>                       
>> a
>>     
>>>>>>>>>>>                       
>>>>>> USB
>>>>>>
>>>>>>
>>>>>>             
>>>>>>>>>>> hard-disk that contains the filesystems.
>>>>>>>>>>>
>>>>>>>>>>> The driver compiles fine on the standard debian etch kernel, and
>>>>>>>>>>>
>>>>>>>>>>>                       
>>>> loads
>>>>
>>>>         
>>>>>>>>>>> perfectly.
>>>>>>>>>>> I just did a minor patch of this driver around a udelay(2500); ->
>>>>>>>>>>> udelay(2000);udelay(500); (2500 is a too big value for the ARM
>>>>>>>>>>> architecture, see bad_udelay).
>>>>>>>>>>>
>>>>>>>>>>> When I try tzap to activate the dvb-t stream, I get those errors:
>>>>>>>>>>> ehci_hcd 0000:00:01.2: alloc_safe_buffer: could not alloc dma
>>>>>>>>>>>                       
>> memory
>>     
>>>>>>>>>>> (size=36096)
>>>>>>>>>>> ehci_hcd 0000:00:01.2: map_single: unable to map unsafe buffer
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>                       
>>>>>> ffc2c000!
>>>>>>
>>>>>>
>>>>>>             
>>>>>>>>>>> many times (I think one per USB packet).
>>>>>>>>>>> And the tuner does not stream anything...
>>>>>>>>>>>
>>>>>>>>>>> I did not modify anything in the NSLU2 (neither RAM or frequency),
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>                       
>>>>>> just
>>>>>>
>>>>>>
>>>>>>             
>>>>>>>>>>> upgraded to debian etch arm architecture following the
>>>>>>>>>>>
>>>>>>>>>>>                       
>>>> nslu2-linux.org
>>>>
>>>>         
>>>>>>>>>>> tutorials.
>>>>>>>>>>> Do you know what it means and how to solve the issue?
>>>>>>>>>>> Do not hesitate to ask for specific information.
>>>>>>>>>>>
>>>>>>>>>>> Thanks!
>>>>>>>>>>> Cheers,
>>>>>>>>>>> David.
>>>>>>>>>>> [email protected]
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> _______________________________________________
>>>>>>>>>>> Em28xx mailing list
>>>>>>>>>>> [email protected]
>>>>>>>>>>> http://mcentral.de/mailman/listinfo/em28xx
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>                       
>>>>>>>>>> I have some informations that can help you to resolve my problem.
>>>>>>>>>>
>>>>>>>>>>                     
>>>> These
>>>>
>>>>         
>>>>>>>>>> informations comes from the newlist of debian-arm :
>>>>>>>>>> In this forum, I said that the driver works correctly on a PC with
>>>>>>>>>> ubuntu, and the response is :
>>>>>>>>>> "Working on a PC does not mean the driver is correct.  On a PC
>>>>>>>>>>
>>>>>>>>>>                     
>>>> physical
>>>>
>>>>         
>>>>>>>>>> addresses match the kernel's point of view in general, while on the
>>>>>>>>>>
>>>>>>>>>>                     
>>>> arm
>>>>
>>>>         
>>>>>>>>>> everything goes through interesting mappings, which make is
>>>>>>>>>>                     
>> essential
>>     
>>>>>>>>>> for the driver to correctly use readX/writeX calls along with
>>>>>>>>>>                     
>> ioremap
>>     
>>>>>>>>>> calls.  Any attemps to directly access anything will fail on the
>>>>>>>>>>                     
>> arm
>>     
>>>>>>>>>> while it almost always works on a PC.  Not sure quite how that
>>>>>>>>>>                     
>> might
>>     
>>>>>>>>>> impact a USB device, but even there the driver has to do DMA setup
>>>>>>>>>>
>>>>>>>>>>                     
>>>> with
>>>>
>>>>         
>>>>>>>>>> corrected addresses, which again is not necesary on a PC but is on
>>>>>>>>>>
>>>>>>>>>>                     
>>>> the
>>>>
>>>>         
>>>>>>>>>> arm. "
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>                     
>>>>>>>>> i'm at the suse conference this week there are some interesting
>>>>>>>>>                   
>> people
>>     
>>>>>>>>> around who're into the usb subsystem.
>>>>>>>>> The em28xx driver makes use of the usb subsystem (which also
>>>>>>>>>                   
>> includes
>>     
>>>>>>>>> allocation of memory)
>>>>>>>>> I'll try to get some answers tomorrow about this issue.
>>>>>>>>>
>>>>>>>>> Markus
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>                   
>>>>>>>> Hello Markus,
>>>>>>>>
>>>>>>>> Do you have news about my problem ?
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>                 
>>>>>>> in em28xx.h there's a static defined value for the number of
>>>>>>>
>>>>>>>               
>>>> videopackets
>>>>
>>>>         
>>>>>>> #define EM28XX_NUM_PACKETS 40
>>>>>>>
>>>>>>> try to lower that value by about the half or less this will allocate
>>>>>>> less memory in the end. I'm not sure if that might out .. although
>>>>>>> that was the only idea which came up after the talk yesterday. Let me
>>>>>>> know if it works out otherwise I'll have a further discussion about it
>>>>>>> with the usb guys.
>>>>>>>
>>>>>>> Markus
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>               
>>>>>> Hello Markus,
>>>>>>
>>>>>> I tried your patch. I set the value to 20. Now tzap seems to work
>>>>>> correctly (tzap -r -c channels.conf "TF1", for exemple) and I don't
>>>>>>             
>> have
>>     
>>>>>> any error with dmesg. But, when i try cat / /dev/dvb/adapter0/dvr0 >
>>>>>> test.ts/, the file is still empty ? and with cp /dev/dvb/adapter0/dvr0
>>>>>> foo, the command doesn't leave immediatly, I must break the command to
>>>>>> leave ?
>>>>>> Do you have an idea ?
>>>>>>
>>>>>>
>>>>>>             
>>>>> Could you try to dump the stream with mplayer?
>>>>> If this doesn't work out we'll add some debug for the usb packets. I
>>>>> only have 2 days left to discuss this issue with the right guys here.
>>>>>
>>>>> Markus
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>           
>>>> Sorry, but it is impossible to use mplayer on a nslu2. It doesn't have a
>>>> video card. I have only a ssh connection.
>>>> You can see some informations about it here : http://www.nslu2-linux.org/
>>>> For information, the scan command doesn't work too.
>>>>
>>>>
>>>>         
>>> The scan command is sufficient.
>>> Could you load the em28xx driver with reg_debug=1 and try to scan for
>>> channels with scan -vvv <frequencymap>
>>>
>>> Thanks,
>>> Markus
>>>
>>>
>>>
>>>       
>> With the scan command, it doesn't find a valid list (I try it on ubuntu
>> and it works) :
>> For example :
>>  >>> tune to:
>> 482000000:INVERSION_AUTO:BANDWIDTH_8_MHZ:FEC_AUTO:FEC_AUTO:QAM_64:TRANSMISSION_MODE_8K:GUARD_INTERVAL_AUTO:HIERARCHY_NONE
>> WARNING: filter timeout pid 0x0011
>> WARNING: filter timeout pid 0x0000
>> WARNING: filter timeout pid 0x0010
>> But no result ?
>>
>> See dmesg informations :
>> DIGITAL TV REQUEST
>> scan: page allocation failure. order:4, mode:0x1
>> Mem-info:
>> DMA per-cpu:
>> cpu 0 hot: high 6, batch 1 used:5
>> cpu 0 cold: high 2, batch 1 used:0
>> DMA32 per-cpu: empty
>> Normal per-cpu: empty
>> HighMem per-cpu: empty
>> Free pages:         740kB (0kB HighMem)
>> Active:4051 inactive:1007 dirty:279 writeback:0 unstable:0 free:185
>> slab:913 map     ped:1468 pagetables:173
>> DMA free:740kB min:724kB low:904kB high:1084kB active:16204kB
>> inactive:4028kB pr     esent:32768kB pages_scanned:0 all_unreclaimable? no
>> lowmem_reserve[]: 0 0 0 0
>> DMA32 free:0kB min:0kB low:0kB high:0kB active:0kB inactive:0kB
>> present:0kB page     s_scanned:0 all_unreclaimable? no
>> lowmem_reserve[]: 0 0 0 0
>> Normal free:0kB min:0kB low:0kB high:0kB active:0kB inactive:0kB
>> present:0kB pag     es_scanned:0 all_unreclaimable? no
>> lowmem_reserve[]: 0 0 0 0
>> HighMem free:0kB min:128kB low:128kB high:128kB active:0kB inactive:0kB
>> present:     0kB pages_scanned:0 all_unreclaimable? no
>> lowmem_reserve[]: 0 0 0 0
>> DMA: 7*4kB 39*8kB 11*16kB 5*32kB 1*64kB 0*128kB 0*256kB 0*512kB 0*1024kB
>> 0*2048k     B 0*4096kB = 740kB
>> DMA32: empty
>> Normal: empty
>> HighMem: empty
>> Swap cache: add 2332, delete 2173, find 267/374, race 0+0
>> Free swap  = 90764kB
>> Total swap = 96348kB
>> Free swap:        90764kB
>> 8192 pages of RAM
>> 356 free pages
>> 753 reserved pages
>> 913 slab pages
>> 7013 pages shared
>> 159 pages swap cached
>>
>> You have enought informations ?
>>
>>     
>
> Could you try to lower the packet number again, seems like you're out
> of DMA'able memory?
>
> I added linux-usb-devel and Greg, the history of that mail can be found at:
> http://mcentral.de/pipermail/em28xx/2007-September/000755.html
>
> Markus
>
>
>   

Do you receive all my informations ?
You can find them at : 
http://mcentral.de/pipermail/em28xx/2007-September/000809.html  (it is a 
rar file)
For information, with the constante set to 10, I still have the same 
problem : "scan: page allocation failure. order:4, mode:0x1"

When I compare with the driver of cinergyT2 that works on NSLU2 (arm 
architecture), I found that the call of the function usb_alloc_urb is 
different :
cinergyT2 : usb_alloc_urb(0, GFP_ATOMIC)
em28xx :  usb_alloc_urb(EM28XX_NUM_PACKETS, GFP_KERNEL);
It is normal ? i am not an expert in driver, it was just a remark ;-)

Thanks
David





-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
[email protected]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel