MMC: CRC Errors with 2GB cards
Hi,
I am seeing a lot of CRC errors happening with CMD25 on a couple of Transcend 2GB cards (specifiation version 1.1). However, if I apply the following patch, I see no errors being reported
drivers/mmc/card/block.c +3 -2
@@ -298,6 +298,7 @@ }if (rq_data_dir (req) !=READ ) { +/* Wait for the card to to be in transfer state*/ do { int err;
@@ -310,8 +311,8 @@ req->rq_disk->disk_name, err); goto cmd_err; } - }while (!(cmd.resp[0 ] &R1_READY_FOR_DATA )); - + }while (!(R1_CURRENT_STATE (cmd.resp[0 ]) ==4 )); +#if 0 if (cmd.resp[0 ] & ~0x00000900 ) printk(KERN_ERR "%s: status = %08x\n",
I noticed that at times the card is not actually in transfer state when it recieves the second consecutive MULTIPLE_BLOCK_WRITE cmd. As per the simplified specification available, all class 4 commands including CMD25 can be issued only in tra state, though its also specified that READY_FOR_DATA indicates whether the card can accept new data.
I am just wondering if I have set of broken cards. Any pointers in this regard?
Thanks and Regards,
Roopesh.
Hi Roopesh,
I'm terribly sorry I haven't replied sooner. I put your message on my todo list, figuring I could get to it within a day or two, but as you noticed, that didn't quite happen. :/
Quoted 6 lines · On Fri, 19 Oct 2007 12:17:00 -0400 Roopesh <[email protected]> wrote:
Hi,
I am seeing a lot of CRC errors happening with CMD25 on a couple of Transcend 2GB cards (specifiation version 1.1). However, if I apply the following patch, I see no errors being reported
Ok, so the card is mishandling the state bits a bit. We should be able to work around that.
drivers/mmc/card/block.c line 316
Quoted 9 lines
drivers/mmc/card/block.c +2
-2@@ -310,8 +311,8 @@ req->rq_disk->disk_name, err); goto cmd_err; } - }while (!(cmd.resp[0 ] &R1_READY_FOR_DATA )); - + }while (!(R1_CURRENT_STATE (cmd.resp[0 ]) ==4 )); +#if 0
This might trade us one problem for another. Could you try the following:
} while (!(cmd.resp[0] & R1_READY_FOR_DATA) || (R1_CURRENT_STATE(cmd.resp[0]) == 7));
This should also work for cards that mishandle the state field, and if the card jumps to some odd state for some reason.
Review line 318
Quoted 4 lines
(quoted diff) +0
-0I am just wondering if I have set of broken cards. Any pointers in this regard?
You do. But don't be too alarmed, that is unfortunately the norm. :/
Rgds
Signature · 5 lines
-- Pierre Ossman Linux kernel, MMC maintainer http://www.kernel.org PulseAudio, core developer http://pulseaudio.org rdesktop, core developer http://www.rdesktop.org
Hi Pierrie,
Thanks for writing back.
: > - } while (!(cmd.resp[0] & R1_READY_FOR_DATA)); : > - : > + } while (!(R1_CURRENT_STATE(cmd.resp[0]) == 4)); : > + : > #if 0
:
: This might trade us one problem for another. Could you try the following:
:
: } while (!(cmd.resp[0] & R1_READY_FOR_DATA) || (R1_CURRENT_STATE(cmd.resp[0]) == 7));
:
:
: This should also work for cards that mishandle the state field, and if the card jumps to
: some odd state for some reason.
:
: > : > I am just wondering if I have set of broken cards. Any pointers : > in this regard? : >
:
: You do. But don't be too alarmed, that is unfortunately the norm. :/
:
I have purchased 4 Transcend cards, and all of them behave the same way. Just wondering why others dont have this problem, and why such a fix hasn't made it to main tree yet.
I have tried the fix, and it works fine. Thanks.
Roopesh.
Quoted 5 lines · On Mon, 29 Oct 2007 05:50:15 -0400 Roopesh <[email protected]> wrote:
I have purchased 4 Transcend cards, and all of them behave the same way. Just wondering why others dont have this problem, and why such a fix hasn't made it to main tree yet.
Odd. I have a whole bunch of transcend cards, and I've never gotten this problem. Perhaps yours are rather recent?
Quoted 2 lines
I have tried the fix, and it works fine. Thanks.
Great. I'll queue it up for -mm for some more testing, after which it will get merged.
Rgds
Signature · 5 lines
-- Pierre Ossman Linux kernel, MMC maintainer http://www.kernel.org PulseAudio, core developer http://pulseaudio.org rdesktop, core developer http://www.rdesktop.org
:---------- Original Message ----------
:From: Pierre Ossman [[email protected]]
:Sent: Wed, Oct 31, 2007 at 03:55:58PM EDT
:
: Odd. I have a whole bunch of transcend cards, and I've
: never gotten this problem. Perhaps yours are rather recent?
:
Though these cards were purchased only in last month, the manufactring date (in cid) is shown to be 3-2007. Here are the entire CID fields:
manfid:28, product_name:SDC , serial=5332, oemid=21334, month-year=3-2007, hw-fw_rev=1-0
/ # cat /sys/bus/mmc/devices/mmc0\:b368/cid
1c5356534443202010000014d400737d
Thanks and Regards,
Roopesh Keeppattu.