* SD card experts wanted @ 2012-08-29 21:17 Juergen Beisert 2012-08-30 3:27 ` Jean-Christophe PLAGNIOL-VILLARD 2012-08-30 10:04 ` Johannes Stezenbach 0 siblings, 2 replies; 6+ messages in thread From: Juergen Beisert @ 2012-08-29 21:17 UTC (permalink / raw) To: barebox Hi there, any SD/MMC card experts here? I have trouble with the current SD/MMC detection code in Barebox. The function sd_change_freq() in the file 'drivers/mci/mci-core.c' tries to detect the max. transfer frequency the attached card can handle. Nothing special here, but the following line in this routine fails on my system: if ((__be32_to_cpu(switch_status[4]) & 0x0f000000) == 0x01000000) It checks the clock speed capabilities returned from the card. As far as I understand the "simplified SD card spec" the "0x01000000" means SDR25 speed, which means still "single transfer per clock at 25 MHz clock rate". But the sd_change_freq() routine sets the MMC_MODE_HS bit for this case in the card's capabilities variable. The remaining routines in 'drivers/mci/mci-core.c' then use 50 MHz for the transfer speed. This behaviour became an issue here, as my SDHC supports 50 MHz, but the attached card only up to 25 MHz. After switching to 50 MHz, the card wasn't able to respond anymore. For SDHCs which supports only up to 25 MHz this misinterpretation is no issue, as the routines still use the 25 MHz, and the SD card can still respond. "simplified SD card spec" talks about the bit's meaning: seems the regular speed was SDR12 which means 12.5 MHz. Then SDR25 cards appear, which are also called "high speed card". Could the MMC_MODE_HS just mean this "high speed" and using it at 50 MHz is just some kind of typo? Or do MMCs act differently and the code mixes SD and MMC features? Any pointers would help. jbe -- Pengutronix e.K. | Juergen Beisert | Linux Solutions for Science and Industry | http://www.pengutronix.de/ | _______________________________________________ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: SD card experts wanted 2012-08-29 21:17 SD card experts wanted Juergen Beisert @ 2012-08-30 3:27 ` Jean-Christophe PLAGNIOL-VILLARD 2012-08-30 10:04 ` Johannes Stezenbach 1 sibling, 0 replies; 6+ messages in thread From: Jean-Christophe PLAGNIOL-VILLARD @ 2012-08-30 3:27 UTC (permalink / raw) To: Juergen Beisert; +Cc: barebox On 23:17 Wed 29 Aug , Juergen Beisert wrote: > Hi there, > > any SD/MMC card experts here? I have trouble with the current SD/MMC detection > code in Barebox. > The function sd_change_freq() in the file 'drivers/mci/mci-core.c' tries to > detect the max. transfer frequency the attached card can handle. Nothing > special here, but the following line in this routine fails on my system: > > if ((__be32_to_cpu(switch_status[4]) & 0x0f000000) == 0x01000000) > > It checks the clock speed capabilities returned from the card. As far as I > understand the "simplified SD card spec" the "0x01000000" means SDR25 speed, > which means still "single transfer per clock at 25 MHz clock rate". But the > sd_change_freq() routine sets the MMC_MODE_HS bit for this case in the card's > capabilities variable. The remaining routines in 'drivers/mci/mci-core.c' > then use 50 MHz for the transfer speed. > > This behaviour became an issue here, as my SDHC supports 50 MHz, but the > attached card only up to 25 MHz. After switching to 50 MHz, the card wasn't > able to respond anymore. > For SDHCs which supports only up to 25 MHz this misinterpretation is no issue, > as the routines still use the 25 MHz, and the SD card can still respond. > > "simplified SD card spec" talks about the bit's meaning: seems the regular > speed was SDR12 which means 12.5 MHz. Then SDR25 cards appear, which are also > called "high speed card". Could the MMC_MODE_HS just mean this "high speed" > and using it at 50 MHz is just some kind of typo? Or do MMCs act differently > and the code mixes SD and MMC features? I saw it too this isssue the mmc and sd feature are different with different spec so we need different code to interpret it see in the kernel Best Regards, J. _______________________________________________ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: SD card experts wanted 2012-08-29 21:17 SD card experts wanted Juergen Beisert 2012-08-30 3:27 ` Jean-Christophe PLAGNIOL-VILLARD @ 2012-08-30 10:04 ` Johannes Stezenbach 2012-08-30 13:08 ` Juergen Beisert 1 sibling, 1 reply; 6+ messages in thread From: Johannes Stezenbach @ 2012-08-30 10:04 UTC (permalink / raw) To: Juergen Beisert; +Cc: barebox Hi, On Wed, Aug 29, 2012 at 11:17:03PM +0200, Juergen Beisert wrote: > any SD/MMC card experts here? I have trouble with the current SD/MMC detection > code in Barebox. > The function sd_change_freq() in the file 'drivers/mci/mci-core.c' tries to > detect the max. transfer frequency the attached card can handle. Nothing > special here, but the following line in this routine fails on my system: > > if ((__be32_to_cpu(switch_status[4]) & 0x0f000000) == 0x01000000) > > It checks the clock speed capabilities returned from the card. As far as I > understand the "simplified SD card spec" the "0x01000000" means SDR25 speed, I think your analysis is correct. > which means still "single transfer per clock at 25 MHz clock rate". But the > sd_change_freq() routine sets the MMC_MODE_HS bit for this case in the card's > capabilities variable. The remaining routines in 'drivers/mci/mci-core.c' > then use 50 MHz for the transfer speed. IMHO mci_startup_sd() is buggy. > This behaviour became an issue here, as my SDHC supports 50 MHz, but the > attached card only up to 25 MHz. After switching to 50 MHz, the card wasn't > able to respond anymore. > For SDHCs which supports only up to 25 MHz this misinterpretation is no issue, > as the routines still use the 25 MHz, and the SD card can still respond. > > "simplified SD card spec" talks about the bit's meaning: seems the regular > speed was SDR12 which means 12.5 MHz. Then SDR25 cards appear, which are also > called "high speed card". Could the MMC_MODE_HS just mean this "high speed" > and using it at 50 MHz is just some kind of typo? Or do MMCs act differently > and the code mixes SD and MMC features? > > Any pointers would help. MMC has the following speed modes: - legacy 0-26MHz - high speed SDR 0-52MHz - high speed DDR 0-52MHz - HS200 SDR 0-200MHz while SD has: - SDR12 - SDR25 - SDR50 - SDR104 - DDR50 So my guess is that MMC_MODE_HS maps to SDR25 for SD-only, and MMC_MODE_HS_52MHz maps to SDR50 for SD and MMC. (Apparently mmc_change_freq always sets MMC_MODE_HS, it is meaningless for MMC.) Thus, mci_startup_sd() should do: if (mci->card_caps & MMC_MODE_HS) mci_set_clock(mci, 25000000); else mci_set_clock(mci, 12500000); Johannes _______________________________________________ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: SD card experts wanted 2012-08-30 10:04 ` Johannes Stezenbach @ 2012-08-30 13:08 ` Juergen Beisert 2012-08-30 16:47 ` Johannes Stezenbach 0 siblings, 1 reply; 6+ messages in thread From: Juergen Beisert @ 2012-08-30 13:08 UTC (permalink / raw) To: barebox Hi Johannes, Johannes Stezenbach wrote: > On Wed, Aug 29, 2012 at 11:17:03PM +0200, Juergen Beisert wrote: > > any SD/MMC card experts here? I have trouble with the current SD/MMC > > detection code in Barebox. > > The function sd_change_freq() in the file 'drivers/mci/mci-core.c' tries > > to detect the max. transfer frequency the attached card can handle. > > Nothing special here, but the following line in this routine fails on my > > system: > > > > if ((__be32_to_cpu(switch_status[4]) & 0x0f000000) == 0x01000000) > > > > It checks the clock speed capabilities returned from the card. As far as > > I understand the "simplified SD card spec" the "0x01000000" means SDR25 > > speed, > > I think your analysis is correct. Thanks. > > which means still "single transfer per clock at 25 MHz clock rate". But > > the sd_change_freq() routine sets the MMC_MODE_HS bit for this case in > > the card's capabilities variable. The remaining routines in > > 'drivers/mci/mci-core.c' then use 50 MHz for the transfer speed. > > IMHO mci_startup_sd() is buggy. Ack. > > This behaviour became an issue here, as my SDHC supports 50 MHz, but the > > attached card only up to 25 MHz. After switching to 50 MHz, the card > > wasn't able to respond anymore. > > For SDHCs which supports only up to 25 MHz this misinterpretation is no > > issue, as the routines still use the 25 MHz, and the SD card can still > > respond. > > > > "simplified SD card spec" talks about the bit's meaning: seems the > > regular speed was SDR12 which means 12.5 MHz. Then SDR25 cards appear, > > which are also called "high speed card". Could the MMC_MODE_HS just mean > > this "high speed" and using it at 50 MHz is just some kind of typo? Or do > > MMCs act differently and the code mixes SD and MMC features? > > > > Any pointers would help. > > MMC has the following speed modes: > > - legacy 0-26MHz > - high speed SDR 0-52MHz > - high speed DDR 0-52MHz > - HS200 SDR 0-200MHz Ahh, here come the curious 26 MHz and 52 MHz into the game. Its related to MMC only. > while SD has: > > - SDR12 > - SDR25 > - SDR50 > - SDR104 > - DDR50 > > So my guess is that MMC_MODE_HS maps to SDR25 for SD-only, > and MMC_MODE_HS_52MHz maps to SDR50 for SD and MMC. > > (Apparently mmc_change_freq always sets MMC_MODE_HS, it is > meaningless for MMC.) > > Thus, mci_startup_sd() should do: > if (mci->card_caps & MMC_MODE_HS) > mci_set_clock(mci, 25000000); > else > mci_set_clock(mci, 12500000); As far as I understand the spec, there is a max. speed field in the CSD which tells us the regular max. speed of this card. And the fields from the CSR can overwrite the CSD settings. So, a card which reports 25 MHz in the CSD can still enable 50 MHz (SDR50) in the CSR and the SDHC then can use 50 MHz for the clock. But maybe I'm wrong here. On the other hand: why switching the clock speed inside the SD card? The spec defines, the SDHC can switch the clock speed at any time, at least it must be below or equal to the max. clock speed the card supports. When the specification is called "simplified", how complicated must be the "non-simplified" version of this document... ;) jbe -- Pengutronix e.K. | Juergen Beisert | Linux Solutions for Science and Industry | http://www.pengutronix.de/ | _______________________________________________ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: SD card experts wanted 2012-08-30 13:08 ` Juergen Beisert @ 2012-08-30 16:47 ` Johannes Stezenbach 2012-08-30 18:05 ` Juergen Beisert 0 siblings, 1 reply; 6+ messages in thread From: Johannes Stezenbach @ 2012-08-30 16:47 UTC (permalink / raw) To: Juergen Beisert; +Cc: barebox Hi, On Thu, Aug 30, 2012 at 03:08:19PM +0200, Juergen Beisert wrote: > Johannes Stezenbach wrote: > > > > MMC has the following speed modes: > > > > - legacy 0-26MHz > > - high speed SDR 0-52MHz > > - high speed DDR 0-52MHz > > - HS200 SDR 0-200MHz > > Ahh, here come the curious 26 MHz and 52 MHz into the game. Its related to MMC > only. > > > while SD has: > > > > - SDR12 > > - SDR25 > > - SDR50 > > - SDR104 > > - DDR50 > > > > So my guess is that MMC_MODE_HS maps to SDR25 for SD-only, > > and MMC_MODE_HS_52MHz maps to SDR50 for SD and MMC. > > > > (Apparently mmc_change_freq always sets MMC_MODE_HS, it is > > meaningless for MMC.) > > > > Thus, mci_startup_sd() should do: > > if (mci->card_caps & MMC_MODE_HS) > > mci_set_clock(mci, 25000000); > > else > > mci_set_clock(mci, 12500000); > > As far as I understand the spec, there is a max. speed field in the CSD which > tells us the regular max. speed of this card. And the fields from the CSR can > overwrite the CSD settings. So, a card which reports 25 MHz in the CSD can > still enable 50 MHz (SDR50) in the CSR and the SDHC then can use 50 MHz for > the clock. But maybe I'm wrong here. What I wrote is not even correct ;-/ SD has legacy Default Mode and legacy High Speed mode, and "Ultra High Speed" modes. SDR12, SDR25, SDR50, SDR104 and DDR50 are all UHS modes. And Default Mode and SDR12 support up to 12.5MB/s, but the bus clock is up to 25MHz (4 bit parallel). So the code in mci_startup_sd() is correct as is. See table in 3.9.6 Summary of Bus Speed Mode. Regarding CSD, my understanding is that TRAN_SPEED only applies to legacy Default Mode. If the card supports HS mode or UHS modes then it must be able to use 50MHz clock. (This is stated below Table 5-6: Maximum Data Transfer Rate Definition) > On the other hand: why switching the clock speed inside the SD card? The spec > defines, the SDHC can switch the clock speed at any time, at least it must be > below or equal to the max. clock speed the card supports. I suppose the card's controller needs to know how to program internal clock dividers. Best Regards Johannes _______________________________________________ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: SD card experts wanted 2012-08-30 16:47 ` Johannes Stezenbach @ 2012-08-30 18:05 ` Juergen Beisert 0 siblings, 0 replies; 6+ messages in thread From: Juergen Beisert @ 2012-08-30 18:05 UTC (permalink / raw) To: barebox Hi Johannes, Johannes Stezenbach wrote: > On Thu, Aug 30, 2012 at 03:08:19PM +0200, Juergen Beisert wrote: > > Johannes Stezenbach wrote: > > > MMC has the following speed modes: > > > > > > - legacy 0-26MHz > > > - high speed SDR 0-52MHz > > > - high speed DDR 0-52MHz > > > - HS200 SDR 0-200MHz > > > > Ahh, here come the curious 26 MHz and 52 MHz into the game. Its related > > to MMC only. > > > > > while SD has: > > > > > > - SDR12 > > > - SDR25 > > > - SDR50 > > > - SDR104 > > > - DDR50 > > > > > > So my guess is that MMC_MODE_HS maps to SDR25 for SD-only, > > > and MMC_MODE_HS_52MHz maps to SDR50 for SD and MMC. > > > > > > (Apparently mmc_change_freq always sets MMC_MODE_HS, it is > > > meaningless for MMC.) > > > > > > Thus, mci_startup_sd() should do: > > > if (mci->card_caps & MMC_MODE_HS) > > > mci_set_clock(mci, 25000000); > > > else > > > mci_set_clock(mci, 12500000); > > > > As far as I understand the spec, there is a max. speed field in the CSD > > which tells us the regular max. speed of this card. And the fields from > > the CSR can overwrite the CSD settings. So, a card which reports 25 MHz > > in the CSD can still enable 50 MHz (SDR50) in the CSR and the SDHC then > > can use 50 MHz for the clock. But maybe I'm wrong here. > > What I wrote is not even correct ;-/ > > SD has legacy Default Mode and legacy High Speed mode, > and "Ultra High Speed" modes. SDR12, SDR25, SDR50, SDR104 and DDR50 > are all UHS modes. And Default Mode and SDR12 support up to 12.5MB/s, > but the bus clock is up to 25MHz (4 bit parallel). > So the code in mci_startup_sd() is correct as is. Now I found the Abbreviations in the spec: UHS Ultra High Speed SDR12 One of UHS modes with single data rate. Up to 12.5MB/sec at 25MHz SDR25 One of UHS modes with single data rate. Up to 25MB/sec at 50MHz SDR50 One of UHS modes with single data rate. Up to 50MB/sec at 100MHz SDR104 One of UHS modes with single data rate. Up to 104MB/sec at 208MHz So they mean the transfer capacity in "bytes per second" and not the clock speed with their SDR<number>. Thanks for this hint. Reading further discovers: * all these UHS modes are _not_ valid for SDSC type of SD cards (up to 2 GiB capacity) This information should help to get the code right. > Regarding CSD, my understanding is that TRAN_SPEED only > applies to legacy Default Mode. If the card supports HS mode > or UHS modes then it must be able to use 50MHz clock. Ack. Regards, Juergen -- Pengutronix e.K. | Juergen Beisert | Linux Solutions for Science and Industry | http://www.pengutronix.de/ | _______________________________________________ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox ^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2012-08-30 18:07 UTC | newest] Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2012-08-29 21:17 SD card experts wanted Juergen Beisert 2012-08-30 3:27 ` Jean-Christophe PLAGNIOL-VILLARD 2012-08-30 10:04 ` Johannes Stezenbach 2012-08-30 13:08 ` Juergen Beisert 2012-08-30 16:47 ` Johannes Stezenbach 2012-08-30 18:05 ` Juergen Beisert
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox