* Query regarding ADMA descriptor allocation
@ 2013-01-11 12:31 Asutosh Das
2013-01-13 12:20 ` Asutosh Das
0 siblings, 1 reply; 3+ messages in thread
From: Asutosh Das @ 2013-01-11 12:31 UTC (permalink / raw)
To: drzeus, linux-mmc
Hi
In sdhci_add_host function, the number of adma descriptors allocated are
128 and each descriptor consumes 8 bytes and 1 4byte is added in the end
like,
kmalloc((128 * 2 + 1) * 4, GFP_KERNEL);
In sdhci_adma_table_pre function, each node in sg-list is iterated and
in case of a misaligned node, a descriptor is consumed, which means for
a single sg-node 2 descriptors were consumed. Let's say in a situation,
where there are multiple unaligned addresses in multiple sg nodes and
the sg-list length is 128, wouldn't this use up all the 128 descriptors
long before the entire sg-list is traversed ?
--
Sent by a consultant of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Query regarding ADMA descriptor allocation
2013-01-11 12:31 Query regarding ADMA descriptor allocation Asutosh Das
@ 2013-01-13 12:20 ` Asutosh Das
2013-01-15 15:16 ` Subhash Jadavani
0 siblings, 1 reply; 3+ messages in thread
From: Asutosh Das @ 2013-01-13 12:20 UTC (permalink / raw)
To: drzeus, linux-mmc
Hi
Can you please let me know if my understanding in the below question is
correct ?
Thanks
Asutosh
On 1/11/2013 6:01 PM, Asutosh Das wrote:
> Hi
> In sdhci_add_host function, the number of adma descriptors allocated
> are 128 and each descriptor consumes 8 bytes and 1 4byte is added in
> the end like,
> kmalloc((128 * 2 + 1) * 4, GFP_KERNEL);
>
> In sdhci_adma_table_pre function, each node in sg-list is iterated and
> in case of a misaligned node, a descriptor is consumed, which means
> for a single sg-node 2 descriptors were consumed. Let's say in a
> situation, where there are multiple unaligned addresses in multiple sg
> nodes and the sg-list length is 128, wouldn't this use up all the 128
> descriptors long before the entire sg-list is traversed ?
>
>
>
>
--
Sent by a consultant of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Query regarding ADMA descriptor allocation
2013-01-13 12:20 ` Asutosh Das
@ 2013-01-15 15:16 ` Subhash Jadavani
0 siblings, 0 replies; 3+ messages in thread
From: Subhash Jadavani @ 2013-01-15 15:16 UTC (permalink / raw)
To: Asutosh Das, Chris Ball; +Cc: drzeus, linux-mmc
Hi Chris,
Can anyone from SDHCi confirm if the understanding below is correct or
not? If it is than we can post a patch to fix it.
Regards,
Subhash
On 1/13/2013 5:50 PM, Asutosh Das wrote:
> Hi
> Can you please let me know if my understanding in the below question
> is correct ?
>
> Thanks
> Asutosh
>
> On 1/11/2013 6:01 PM, Asutosh Das wrote:
>> Hi
>> In sdhci_add_host function, the number of adma descriptors allocated
>> are 128 and each descriptor consumes 8 bytes and 1 4byte is added in
>> the end like,
>> kmalloc((128 * 2 + 1) * 4, GFP_KERNEL);
>>
>> In sdhci_adma_table_pre function, each node in sg-list is iterated
>> and in case of a misaligned node, a descriptor is consumed, which
>> means for a single sg-node 2 descriptors were consumed. Let's say in
>> a situation, where there are multiple unaligned addresses in multiple
>> sg nodes and the sg-list length is 128, wouldn't this use up all the
>> 128 descriptors long before the entire sg-list is traversed ?
>>
>>
>>
>>
>
>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2013-01-15 15:16 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-01-11 12:31 Query regarding ADMA descriptor allocation Asutosh Das
2013-01-13 12:20 ` Asutosh Das
2013-01-15 15:16 ` Subhash Jadavani
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).