(Linux3.10.17)driver/mmc/host/sdhci.c throw warnings

[xlin]

Greetings,

I am working on a SDIO device driver on imx6sl EVK.  I am running Linux3.10.17_ga.

I run into a warning from driver/mmc/host/sdhci.c  occasionally.  I would appreciate if you would like to share with me on whether you have seen the same warning or you have an idea about what may cause this warning.

Best Regards,

Xinsi Lin 

------------[ cut here ]------------

WARNING: at drivers/mmc/host/sdhci.c:854 sdhci_send_command+0xdf0/0xeb4()

Modules linked in: dhd(O) evbug

CPU: 0 PID: 633 Comm: dhd_dpc Tainted: G        W  O 3.10.17-1.0.0_ga+g232293e #30

[<80013ae0>] (unwind_backtrace+0x0/0xf4) from [<80011504>] (show_stack+0x10/0x14)

[<80011504>] (show_stack+0x10/0x14) from [<800260d0>] (warn_slowpath_common+0x54/0x6c)

[<800260d0>] (warn_slowpath_common+0x54/0x6c) from [<80026184>] (warn_slowpath_null+0x1c/0x24)

[<80026184>] (warn_slowpath_null+0x1c/0x24) from [<80452820>] (sdhci_send_command+0xdf0/0xeb4)

[<80452820>] (sdhci_send_command+0xdf0/0xeb4) from [<80454014>] (sdhci_request+0xc0/0x1ec)

[<80454014>] (sdhci_request+0xc0/0x1ec) from [<8043f920>] (mmc_start_request+0xc4/0xe0)

[<8043f920>] (mmc_start_request+0xc4/0xe0) from [<8043fa54>] (__mmc_start_req+0x60/0x84)

[<8043fa54>] (__mmc_start_req+0x60/0x84) from [<8043fafc>] (mmc_wait_for_req+0x10/0x20)

[<8043fafc>] (mmc_wait_for_req+0x10/0x20) from [<7f024a90>] (sdioh_request_packet_chain+0x31c/0x45c [dhd])

[<7f024a90>] (sdioh_request_packet_chain+0x31c/0x45c [dhd]) from [<7f025b40>] (sdioh_request_buffer+0x118/0x2a8 [dhd])

[<7f025b40>] (sdioh_request_buffer+0x118/0x2a8 [dhd]) from [<7f026908>] (bcmsdh_send_buf+0x84/0x98 [dhd])

[<7f026908>] (bcmsdh_send_buf+0x84/0x98 [dhd]) from [<7f01f1f4>] (dhd_bcmsdh_send_buf.constprop.26+0x64/0x1b0 [dhd])

[<7f01f1f4>] (dhd_bcmsdh_send_buf.constprop.26+0x64/0x1b0 [dhd]) from [<7f01f9b0>] (dhdsdio_txpkt.constprop.25+0x5f0/0x688 [dhd])

[<7f01f9b0>] (dhdsdio_txpkt.constprop.25+0x5f0/0x688 [dhd]) from [<7f01fb68>] (dhdsdio_sendfromq+0x120/0x330 [dhd])

[<7f01fb68>] (dhdsdio_sendfromq+0x120/0x330 [dhd]) from [<7f022d18>] (dhd_bus_dpc+0x5b4/0xc80 [dhd])

[<7f022d18>] (dhd_bus_dpc+0x5b4/0xc80 [dhd]) from [<7f0071c0>] (dhd_dpc_thread+0xd4/0xec [dhd])

[<7f0071c0>] (dhd_dpc_thread+0xd4/0xec [dhd]) from [<8004345c>] (kthread+0xb4/0xb8)

[<8004345c>] (kthread+0xb4/0xb8) from [<8000e118>] (ret_from_fork+0x14/0x3c)

---[ end trace 3a302636f037804c ]---

[igorpadykov]

Hi Xinsi

these warnings are caused by inaccuracies in driver porting, please look below,

so recommended to consult with vendor of that device of such kind of messages

http://e2e.ti.com/support/wireless_connectivity/f/307/p/356133/1250633

Best regards

igor

-----------------------------------------------------------------------------------------------------------------------

Note: If this post answers your question, please click the Correct Answer button. Thank you!

-----------------------------------------------------------------------------------------------------------------------

[xlin]

Hi Igor,

Thank you for your reply. I took a look at the link you provided but it is not the same issue.  I took a close look at the code that throw the warning. In the code except below, on the line 842(in my code the same line of code is on line 854), it warns about "sg_cnt!=1". Here I have a question about  why there is  the restriction of "sg_cnt==1" by the sdhci driver.  The im6sl DMA control supports chained sg buffers, right?   

826                 } else {

827                         int sg_cnt;

828

829                         sg_cnt = dma_map_sg(mmc_dev(host->mmc),

830                                         data->sg, data->sg_len,

831  (data->flags & MMC_DATA_READ) ?

832                                                 DMA_FROM_DEVICE :

833                                                 DMA_TO_DEVICE);

834                         if (sg_cnt == 0) {

835                                 /*

836                                  * This only happens when someone fed

837                                  * us an invalid request.

838                                  */

839                                 WARN_ON(1);

840                                 host->flags &= ~SDHCI_REQ_USE_DMA;

841                         } else {

842                                 WARN_ON(sg_cnt != 1);

843                                 sdhci_writel(host, sg_dma_address(data->sg),

844                                         SDHCI_DMA_ADDRESS);

845                         }

846                 }

Thanks,

Xinsi Lin

[KursadOney]

Xinsi,

The warning in line 842 is in the core driver. It's from upstream linux, it's not an FSL addition. The warning is there because the code that follows it processes only a single dma request:

sdhci_writel(host, sg_dma_address(data->sg),

        SDHCI_DMA_ADDRESS);

If your driver is returning more than 1 sg, you will need to modify this code to iterate through all of them using the for_each_sg macro. Note that there might be side effects to this. You'd be better off discussing it with the maintainer of the core driver or in its relevant mailing list.

[igorpadykov]

Hi Xinsi

do you have these warnings with unchanged Demo codes ?

L3.10.17_1.0.2_iMX6SL_BUNDLE : i.MX 6Sololite, i.MXSoloEVK Linux Binary Demo

Best regards

igor

[xlin]

Hi Igor,

Yes, I have seen the warning with the kernel image I built from the original untouched source code. On the current image that I used to report the issue,  actually I didn't change any logic in the code , I just added some debugging prints.

Cheers,

Xinsi

[igorpadykov]

Hi Xinsi

nevertheless do you have these warnings with prebuilt  Demo image,

without building?

L3.10.17_1.0.2_iMX6SL_BUNDLE : i.MX 6Sololite, i.MXSoloEVK Linux Binary Demo

Best regards

igor

[xlin]

I can't do insmod my driver with the pre-build image. the driver has to match the symbol version of kernel. Except for the symbol version,  I believe my build image shall be identical logical wise to your pre-built.

Xinsi

[igorpadykov]

Hi Xinsi

so, seems warnings are caused by your driver changes.

Best regards

igor

[xlin]

Hi Igor,

Yes or not.  Your SDIO driver is not exercised without my driver. I said earlier in the thread the warning  was against  using chained sg-buffer. I asked the question before-"Here I have a question about  why there is  the restriction of "sg_cnt==1" by the sdhci driver.  The imx6sl DMA control supports chained sg buffers, right?". Based on imx6sl reference manual, the imx6sl DMA operations support chained sg buffers, I am trying to understand why your sdhci driver warns about using chained sg buffers?

Thanks,

Xinsi

[igorpadykov]

Hi Xinsi

what SDIO chip you are using ?

Not sure that it is possible to give comments on every piece of code

just based on warning messages.

Best regards

igor

[xlin]

Igor, I am using the sdhci1 of imx6sl evk. Xinsi 

[igorpadykov]

Hi Xinsi

what SDIO chip is connected to sdhci1 of imx6sl evk ?

~igor

[xlin]

Sorry, Igor.  It is a Broadcom Wifi chipset(4339). Xinsi

[xlin]

want to clarify that I am not using ADMA.