freebsd-dev/sys/dev/mmc/mmcsd.c

1415 lines
38 KiB
C
Raw Normal View History

/*-
* Copyright (c) 2006 Bernd Walter. All rights reserved.
* Copyright (c) 2006 M. Warner Losh. All rights reserved.
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
* Copyright (c) 2017 Marius Strobl <marius@FreeBSD.org>
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* Portions of this software may have been developed with reference to
* the SD Simplified Specification. The following disclaimer may apply:
*
* The following conditions apply to the release of the simplified
* specification ("Simplified Specification") by the SD Card Association and
* the SD Group. The Simplified Specification is a subset of the complete SD
* Specification which is owned by the SD Card Association and the SD
* Group. This Simplified Specification is provided on a non-confidential
* basis subject to the disclaimers below. Any implementation of the
* Simplified Specification may require a license from the SD Card
* Association, SD Group, SD-3C LLC or other third parties.
*
* Disclaimers:
*
* The information contained in the Simplified Specification is presented only
* as a standard specification for SD Cards and SD Host/Ancillary products and
* is provided "AS-IS" without any representations or warranties of any
* kind. No responsibility is assumed by the SD Group, SD-3C LLC or the SD
* Card Association for any damages, any infringements of patents or other
* right of the SD Group, SD-3C LLC, the SD Card Association or any third
* parties, which may result from its use. No license is granted by
* implication, estoppel or otherwise under any patent or other rights of the
* SD Group, SD-3C LLC, the SD Card Association or any third party. Nothing
* herein shall be construed as an obligation by the SD Group, the SD-3C LLC
* or the SD Card Association to disclose or distribute any technical
* information, know-how or other confidential information to any third party.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bio.h>
#include <sys/bus.h>
#include <sys/conf.h>
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
#include <sys/fcntl.h>
#include <sys/ioccom.h>
#include <sys/kernel.h>
#include <sys/kthread.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/mutex.h>
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
#include <sys/slicer.h>
#include <sys/time.h>
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
#include <geom/geom.h>
#include <geom/geom_disk.h>
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
#include <dev/mmc/bridge.h>
#include <dev/mmc/mmc_ioctl.h>
#include <dev/mmc/mmc_subr.h>
#include <dev/mmc/mmcbrvar.h>
#include <dev/mmc/mmcreg.h>
#include <dev/mmc/mmcvar.h>
#include "mmcbus_if.h"
#if __FreeBSD_version < 800002
#define kproc_create kthread_create
#define kproc_exit kthread_exit
#endif
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
#define MMCSD_CMD_RETRIES 5
#define MMCSD_FMT_BOOT "mmcsd%dboot"
#define MMCSD_FMT_GP "mmcsd%dgp"
#define MMCSD_FMT_RPMB "mmcsd%drpmb"
#define MMCSD_LABEL_ENH "enh"
#define MMCSD_PART_NAMELEN (16 + 1)
struct mmcsd_softc;
struct mmcsd_part {
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
struct mtx disk_mtx;
struct mtx ioctl_mtx;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
struct mmcsd_softc *sc;
struct disk *disk;
struct proc *p;
struct bio_queue_head bio_queue;
daddr_t eblock, eend; /* Range remaining after the last erase. */
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
u_int cnt;
u_int type;
int running;
int suspend;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
int ioctl;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
bool ro;
char name[MMCSD_PART_NAMELEN];
};
struct mmcsd_softc {
device_t dev;
device_t mmcbr;
struct mmcsd_part *part[MMC_PART_MAX];
enum mmc_card_mode mode;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
u_int max_data; /* Maximum data size [blocks] */
u_int erase_sector; /* Device native erase sector size [blocks] */
uint8_t high_cap; /* High Capacity device (block addressed) */
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
uint8_t part_curr; /* Partition currently switched to */
uint8_t ext_csd[MMC_EXTCSD_SIZE];
uint16_t rca;
uint32_t part_time; /* Partition switch timeout [us] */
off_t enh_base; /* Enhanced user data area slice base ... */
off_t enh_size; /* ... and size [bytes] */
int log_count;
struct timeval log_time;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
struct cdev *rpmb_dev;
};
static const char *errmsg[] =
{
"None",
"Timeout",
"Bad CRC",
"Fifo",
"Failed",
"Invalid",
"NO MEMORY"
};
#define LOG_PPS 5 /* Log no more than 5 errors per second. */
/* bus entry points */
static int mmcsd_attach(device_t dev);
static int mmcsd_detach(device_t dev);
static int mmcsd_probe(device_t dev);
/* disk routines */
static int mmcsd_close(struct disk *dp);
static int mmcsd_dump(void *arg, void *virtual, vm_offset_t physical,
off_t offset, size_t length);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
static int mmcsd_getattr(struct bio *);
static int mmcsd_ioctl_disk(struct disk *disk, u_long cmd, void *data,
int fflag, struct thread *td);
static int mmcsd_open(struct disk *dp);
static void mmcsd_strategy(struct bio *bp);
static void mmcsd_task(void *arg);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
/* RMPB cdev interface */
static int mmcsd_ioctl_rpmb(struct cdev *dev, u_long cmd, caddr_t data,
int fflag, struct thread *td);
static void mmcsd_add_part(struct mmcsd_softc *sc, u_int type,
const char *name, u_int cnt, off_t media_size, off_t erase_size, bool ro);
static int mmcsd_bus_bit_width(device_t dev);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
static daddr_t mmcsd_delete(struct mmcsd_part *part, struct bio *bp);
static int mmcsd_ioctl(struct mmcsd_part *part, u_long cmd, void *data,
int fflag);
static int mmcsd_ioctl_cmd(struct mmcsd_part *part, struct mmc_ioc_cmd *mic,
int fflag);
static uintmax_t mmcsd_pretty_size(off_t size, char *unit);
static daddr_t mmcsd_rw(struct mmcsd_part *part, struct bio *bp);
static int mmcsd_set_blockcount(struct mmcsd_softc *sc, u_int count, bool rel);
static int mmcsd_slicer(device_t dev, const char *provider,
struct flash_slice *slices, int *nslices);
static int mmcsd_switch_part(device_t bus, device_t dev, uint16_t rca,
u_int part);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
#define MMCSD_DISK_LOCK(_part) mtx_lock(&(_part)->disk_mtx)
#define MMCSD_DISK_UNLOCK(_part) mtx_unlock(&(_part)->disk_mtx)
#define MMCSD_DISK_LOCK_INIT(_part) \
mtx_init(&(_part)->disk_mtx, (_part)->name, "mmcsd disk", MTX_DEF)
#define MMCSD_DISK_LOCK_DESTROY(_part) mtx_destroy(&(_part)->disk_mtx);
#define MMCSD_DISK_ASSERT_LOCKED(_part) \
mtx_assert(&(_part)->disk_mtx, MA_OWNED);
#define MMCSD_DISK_ASSERT_UNLOCKED(_part) \
mtx_assert(&(_part)->disk_mtx, MA_NOTOWNED);
#define MMCSD_IOCTL_LOCK(_part) mtx_lock(&(_part)->ioctl_mtx)
#define MMCSD_IOCTL_UNLOCK(_part) mtx_unlock(&(_part)->ioctl_mtx)
#define MMCSD_IOCTL_LOCK_INIT(_part) \
mtx_init(&(_part)->ioctl_mtx, (_part)->name, "mmcsd IOCTL", MTX_DEF)
#define MMCSD_IOCTL_LOCK_DESTROY(_part) mtx_destroy(&(_part)->ioctl_mtx);
#define MMCSD_IOCTL_ASSERT_LOCKED(_part) \
mtx_assert(&(_part)->ioctl_mtx, MA_OWNED);
#define MMCSD_IOCLT_ASSERT_UNLOCKED(_part) \
mtx_assert(&(_part)->ioctl_mtx, MA_NOTOWNED);
static int
mmcsd_probe(device_t dev)
{
device_quiet(dev);
device_set_desc(dev, "MMC/SD Memory Card");
return (0);
}
static int
mmcsd_attach(device_t dev)
{
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
device_t mmcbr;
struct mmcsd_softc *sc;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
const uint8_t *ext_csd;
off_t erase_size, sector_size, size, wp_size;
uintmax_t bytes;
int err, i;
uint8_t rev;
bool comp, ro;
char unit[2];
sc = device_get_softc(dev);
sc->dev = dev;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
sc->mmcbr = mmcbr = device_get_parent(dev);
sc->mode = mmcbr_get_mode(mmcbr);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
/*
* Note that in principle with an SDHCI-like re-tuning implementation,
* the maximum data size can change at runtime due to a device removal/
* insertion that results in switches to/from a transfer mode involving
* re-tuning, iff there are multiple devices on a given bus. Until now
* mmc(4) lacks support for rescanning already attached buses, however,
* and sdhci(4) has no support for embedded/shared buses in the first
* place either.
*/
sc->max_data = mmc_get_max_data(dev);
sc->erase_sector = mmc_get_erase_sector(dev);
sc->high_cap = mmc_get_high_cap(dev);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
sc->rca = mmc_get_rca(dev);
/* Only MMC >= 4.x devices support EXT_CSD. */
if (mmc_get_spec_vers(dev) >= 4) {
MMCBUS_ACQUIRE_BUS(mmcbr, dev);
err = mmc_send_ext_csd(mmcbr, dev, sc->ext_csd);
MMCBUS_RELEASE_BUS(mmcbr, dev);
if (err != MMC_ERR_NONE)
bzero(sc->ext_csd, sizeof(sc->ext_csd));
}
ext_csd = sc->ext_csd;
/*
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
* Enhanced user data area and general purpose partitions are only
* supported in revision 1.4 (EXT_CSD_REV == 4) and later, the RPMB
* partition in revision 1.5 (MMC v4.41, EXT_CSD_REV == 5) and later.
*/
rev = ext_csd[EXT_CSD_REV];
/*
* Ignore user-creatable enhanced user data area and general purpose
* partitions partitions as long as partitioning hasn't been finished.
*/
comp = (ext_csd[EXT_CSD_PART_SET] & EXT_CSD_PART_SET_COMPLETED) != 0;
/*
* Add enhanced user data area slice, unless it spans the entirety of
* the user data area. The enhanced area is of a multiple of high
* capacity write protect groups ((ERASE_GRP_SIZE + HC_WP_GRP_SIZE) *
* 512 KB) and its offset given in either sectors or bytes, depending
* on whether it's a high capacity device or not.
* NB: The slicer and its slices need to be registered before adding
* the disk for the corresponding user data area as re-tasting is
* racy.
*/
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
sector_size = mmc_get_sector_size(dev);
size = ext_csd[EXT_CSD_ENH_SIZE_MULT] +
(ext_csd[EXT_CSD_ENH_SIZE_MULT + 1] << 8) +
(ext_csd[EXT_CSD_ENH_SIZE_MULT + 2] << 16);
if (rev >= 4 && comp == TRUE && size > 0 &&
(ext_csd[EXT_CSD_PART_SUPPORT] &
EXT_CSD_PART_SUPPORT_ENH_ATTR_EN) != 0 &&
(ext_csd[EXT_CSD_PART_ATTR] & (EXT_CSD_PART_ATTR_ENH_USR)) != 0) {
erase_size = ext_csd[EXT_CSD_ERASE_GRP_SIZE] * 1024 *
MMC_SECTOR_SIZE;
wp_size = ext_csd[EXT_CSD_HC_WP_GRP_SIZE];
size *= erase_size * wp_size;
if (size != mmc_get_media_size(dev) * sector_size) {
sc->enh_size = size;
sc->enh_base = (ext_csd[EXT_CSD_ENH_START_ADDR] +
(ext_csd[EXT_CSD_ENH_START_ADDR + 1] << 8) +
(ext_csd[EXT_CSD_ENH_START_ADDR + 2] << 16) +
(ext_csd[EXT_CSD_ENH_START_ADDR + 3] << 24)) *
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
(sc->high_cap != 0 ? MMC_SECTOR_SIZE : 1);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
} else if (bootverbose)
device_printf(dev,
"enhanced user data area spans entire device\n");
}
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
/*
* Add default partition. This may be the only one or the user
* data area in case partitions are supported.
*/
ro = mmc_get_read_only(dev);
mmcsd_add_part(sc, EXT_CSD_PART_CONFIG_ACC_DEFAULT, "mmcsd",
device_get_unit(dev), mmc_get_media_size(dev) * sector_size,
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
sc->erase_sector * sector_size, ro);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
if (mmc_get_spec_vers(dev) < 3)
return (0);
/* Belatedly announce enhanced user data slice. */
if (sc->enh_size != 0) {
bytes = mmcsd_pretty_size(size, unit);
printf(FLASH_SLICES_FMT ": %ju%sB enhanced user data area "
"slice offset 0x%jx at %s\n", device_get_nameunit(dev),
MMCSD_LABEL_ENH, bytes, unit, (uintmax_t)sc->enh_base,
device_get_nameunit(dev));
}
/*
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
* Determine partition switch timeout (provided in units of 10 ms)
* and ensure it's at least 300 ms as some eMMC chips lie.
*/
sc->part_time = max(ext_csd[EXT_CSD_PART_SWITCH_TO] * 10 * 1000,
300 * 1000);
/* Add boot partitions, which are of a fixed multiple of 128 KB. */
size = ext_csd[EXT_CSD_BOOT_SIZE_MULT] * MMC_BOOT_RPMB_BLOCK_SIZE;
if (size > 0 && (mmcbr_get_caps(mmcbr) & MMC_CAP_BOOT_NOACC) == 0) {
mmcsd_add_part(sc, EXT_CSD_PART_CONFIG_ACC_BOOT0,
MMCSD_FMT_BOOT, 0, size, MMC_BOOT_RPMB_BLOCK_SIZE,
ro | ((ext_csd[EXT_CSD_BOOT_WP_STATUS] &
EXT_CSD_BOOT_WP_STATUS_BOOT0_MASK) != 0));
mmcsd_add_part(sc, EXT_CSD_PART_CONFIG_ACC_BOOT1,
MMCSD_FMT_BOOT, 1, size, MMC_BOOT_RPMB_BLOCK_SIZE,
ro | ((ext_csd[EXT_CSD_BOOT_WP_STATUS] &
EXT_CSD_BOOT_WP_STATUS_BOOT1_MASK) != 0));
}
/* Add RPMB partition, which also is of a fixed multiple of 128 KB. */
size = ext_csd[EXT_CSD_RPMB_MULT] * MMC_BOOT_RPMB_BLOCK_SIZE;
if (rev >= 5 && size > 0)
mmcsd_add_part(sc, EXT_CSD_PART_CONFIG_ACC_RPMB,
MMCSD_FMT_RPMB, 0, size, MMC_BOOT_RPMB_BLOCK_SIZE, ro);
if (rev <= 3 || comp == FALSE)
return (0);
/*
* Add general purpose partitions, which are of a multiple of high
* capacity write protect groups, too.
*/
if ((ext_csd[EXT_CSD_PART_SUPPORT] & EXT_CSD_PART_SUPPORT_EN) != 0) {
erase_size = ext_csd[EXT_CSD_ERASE_GRP_SIZE] * 1024 *
MMC_SECTOR_SIZE;
wp_size = ext_csd[EXT_CSD_HC_WP_GRP_SIZE];
for (i = 0; i < MMC_PART_GP_MAX; i++) {
size = ext_csd[EXT_CSD_GP_SIZE_MULT + i * 3] +
(ext_csd[EXT_CSD_GP_SIZE_MULT + i * 3 + 1] << 8) +
(ext_csd[EXT_CSD_GP_SIZE_MULT + i * 3 + 2] << 16);
if (size == 0)
continue;
mmcsd_add_part(sc, EXT_CSD_PART_CONFIG_ACC_GP0 + i,
MMCSD_FMT_GP, i, size * erase_size * wp_size,
erase_size, ro);
}
}
return (0);
}
static uintmax_t
mmcsd_pretty_size(off_t size, char *unit)
{
uintmax_t bytes;
int i;
/*
* Display in most natural units. There's no card < 1MB. However,
* RPMB partitions occasionally are smaller than that, though. The
* SD standard goes to 2 GiB due to its reliance on FAT, but the data
* format supports up to 4 GiB and some card makers push it up to this
* limit. The SDHC standard only goes to 32 GiB due to FAT32, but the
* data format supports up to 2 TiB however. 2048 GB isn't too ugly,
* so we note it in passing here and don't add the code to print TB).
* Since these cards are sold in terms of MB and GB not MiB and GiB,
* report them like that. We also round to the nearest unit, since
* many cards are a few percent short, even of the power of 10 size.
*/
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
bytes = size;
unit[0] = unit[1] = '\0';
for (i = 0; i <= 2 && bytes >= 1000; i++) {
bytes = (bytes + 1000 / 2 - 1) / 1000;
switch (i) {
case 0:
unit[0] = 'k';
break;
case 1:
unit[0] = 'M';
break;
case 2:
unit[0] = 'G';
break;
default:
break;
}
}
return (bytes);
}
static struct cdevsw mmcsd_rpmb_cdevsw = {
.d_version = D_VERSION,
.d_name = "mmcsdrpmb",
.d_ioctl = mmcsd_ioctl_rpmb
};
static void
mmcsd_add_part(struct mmcsd_softc *sc, u_int type, const char *name, u_int cnt,
off_t media_size, off_t erase_size, bool ro)
{
struct make_dev_args args;
device_t dev, mmcbr;
const char *ext;
const uint8_t *ext_csd;
struct mmcsd_part *part;
struct disk *d;
uintmax_t bytes;
u_int gp;
uint32_t speed;
uint8_t extattr;
bool enh;
char unit[2];
dev = sc->dev;
mmcbr = sc->mmcbr;
part = sc->part[type] = malloc(sizeof(*part), M_DEVBUF,
M_WAITOK | M_ZERO);
part->sc = sc;
part->cnt = cnt;
part->type = type;
part->ro = ro;
snprintf(part->name, sizeof(part->name), name, device_get_unit(dev));
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_IOCTL_LOCK_INIT(part);
/*
* For the RPMB partition, allow IOCTL access only.
* NB: If ever attaching RPMB partitions to disk(9), the re-tuning
* implementation and especially its pausing need to be revisited,
* because then re-tuning requests may be issued by the IOCTL half
* of this driver while re-tuning is already paused by the disk(9)
* one and vice versa.
*/
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
if (type == EXT_CSD_PART_CONFIG_ACC_RPMB) {
make_dev_args_init(&args);
args.mda_flags = MAKEDEV_CHECKNAME | MAKEDEV_WAITOK;
args.mda_devsw = &mmcsd_rpmb_cdevsw;
args.mda_uid = UID_ROOT;
args.mda_gid = GID_OPERATOR;
args.mda_mode = 0640;
args.mda_si_drv1 = part;
if (make_dev_s(&args, &sc->rpmb_dev, "%s", part->name) != 0) {
device_printf(dev, "Failed to make RPMB device\n");
free(part, M_DEVBUF);
return;
}
} else {
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_DISK_LOCK_INIT(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
d = part->disk = disk_alloc();
d->d_open = mmcsd_open;
d->d_close = mmcsd_close;
d->d_strategy = mmcsd_strategy;
d->d_ioctl = mmcsd_ioctl_disk;
d->d_dump = mmcsd_dump;
d->d_getattr = mmcsd_getattr;
d->d_name = part->name;
d->d_drv1 = part;
d->d_sectorsize = mmc_get_sector_size(dev);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
d->d_maxsize = sc->max_data * d->d_sectorsize;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
d->d_mediasize = media_size;
d->d_stripesize = erase_size;
d->d_unit = cnt;
d->d_flags = DISKFLAG_CANDELETE;
d->d_delmaxsize = erase_size;
strlcpy(d->d_ident, mmc_get_card_sn_string(dev),
sizeof(d->d_ident));
strlcpy(d->d_descr, mmc_get_card_id_string(dev),
sizeof(d->d_descr));
d->d_rotation_rate = DISK_RR_NON_ROTATING;
disk_create(d, DISK_VERSION);
bioq_init(&part->bio_queue);
part->running = 1;
kproc_create(&mmcsd_task, part, &part->p, 0, 0,
"%s%d: mmc/sd card", part->name, cnt);
}
bytes = mmcsd_pretty_size(media_size, unit);
if (type == EXT_CSD_PART_CONFIG_ACC_DEFAULT) {
speed = mmcbr_get_clock(mmcbr);
printf("%s%d: %ju%sB <%s>%s at %s %d.%01dMHz/%dbit/%d-block\n",
part->name, cnt, bytes, unit, mmc_get_card_id_string(dev),
ro ? " (read-only)" : "", device_get_nameunit(mmcbr),
speed / 1000000, (speed / 100000) % 10,
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
mmcsd_bus_bit_width(dev), sc->max_data);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
} else if (type == EXT_CSD_PART_CONFIG_ACC_RPMB) {
printf("%s: %ju%sB partion %d%s at %s\n", part->name, bytes,
unit, type, ro ? " (read-only)" : "",
device_get_nameunit(dev));
} else {
enh = false;
ext = NULL;
extattr = 0;
if (type >= EXT_CSD_PART_CONFIG_ACC_GP0 &&
type <= EXT_CSD_PART_CONFIG_ACC_GP3) {
ext_csd = sc->ext_csd;
gp = type - EXT_CSD_PART_CONFIG_ACC_GP0;
if ((ext_csd[EXT_CSD_PART_SUPPORT] &
EXT_CSD_PART_SUPPORT_ENH_ATTR_EN) != 0 &&
(ext_csd[EXT_CSD_PART_ATTR] &
(EXT_CSD_PART_ATTR_ENH_GP0 << gp)) != 0)
enh = true;
else if ((ext_csd[EXT_CSD_PART_SUPPORT] &
EXT_CSD_PART_SUPPORT_EXT_ATTR_EN) != 0) {
extattr = (ext_csd[EXT_CSD_EXT_PART_ATTR +
(gp / 2)] >> (4 * (gp % 2))) & 0xF;
switch (extattr) {
case EXT_CSD_EXT_PART_ATTR_DEFAULT:
break;
case EXT_CSD_EXT_PART_ATTR_SYSTEMCODE:
ext = "system code";
break;
case EXT_CSD_EXT_PART_ATTR_NPERSISTENT:
ext = "non-persistent";
break;
default:
ext = "reserved";
break;
}
}
}
if (ext == NULL)
printf("%s%d: %ju%sB partion %d%s%s at %s\n",
part->name, cnt, bytes, unit, type, enh ?
" enhanced" : "", ro ? " (read-only)" : "",
device_get_nameunit(dev));
else
printf("%s%d: %ju%sB partion %d extended 0x%x "
"(%s)%s at %s\n", part->name, cnt, bytes, unit,
type, extattr, ext, ro ? " (read-only)" : "",
device_get_nameunit(dev));
}
}
static int
mmcsd_slicer(device_t dev, const char *provider,
struct flash_slice *slices, int *nslices)
{
char name[MMCSD_PART_NAMELEN];
struct mmcsd_softc *sc;
struct mmcsd_part *part;
*nslices = 0;
if (slices == NULL)
return (ENOMEM);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
sc = device_get_softc(dev);
if (sc->enh_size == 0)
return (ENXIO);
part = sc->part[EXT_CSD_PART_CONFIG_ACC_DEFAULT];
snprintf(name, sizeof(name), "%s%d", part->disk->d_name,
part->disk->d_unit);
if (strcmp(name, provider) != 0)
return (ENXIO);
*nslices = 1;
slices[0].base = sc->enh_base;
slices[0].size = sc->enh_size;
slices[0].label = MMCSD_LABEL_ENH;
return (0);
}
static int
mmcsd_detach(device_t dev)
{
struct mmcsd_softc *sc = device_get_softc(dev);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
struct mmcsd_part *part;
int i;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
for (i = 0; i < MMC_PART_MAX; i++) {
part = sc->part[i];
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
if (part != NULL) {
if (part->disk != NULL) {
MMCSD_DISK_LOCK(part);
part->suspend = 0;
if (part->running > 0) {
/* kill thread */
part->running = 0;
wakeup(part);
/* wait for thread to finish. */
while (part->running != -1)
msleep(part, &part->disk_mtx, 0,
"mmcsd disk detach", 0);
}
MMCSD_DISK_UNLOCK(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
}
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_IOCTL_LOCK(part);
while (part->ioctl > 0)
msleep(part, &part->ioctl_mtx, 0,
"mmcsd IOCTL detach", 0);
part->ioctl = -1;
MMCSD_IOCTL_UNLOCK(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
}
}
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
if (sc->rpmb_dev != NULL)
destroy_dev(sc->rpmb_dev);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
for (i = 0; i < MMC_PART_MAX; i++) {
part = sc->part[i];
if (part != NULL) {
if (part->disk != NULL) {
/* Flush the request queue. */
bioq_flush(&part->bio_queue, NULL, ENXIO);
/* kill disk */
disk_destroy(part->disk);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_DISK_LOCK_DESTROY(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
}
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_IOCTL_LOCK_DESTROY(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
free(part, M_DEVBUF);
}
}
return (0);
}
static int
mmcsd_suspend(device_t dev)
{
struct mmcsd_softc *sc = device_get_softc(dev);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
struct mmcsd_part *part;
int i;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
for (i = 0; i < MMC_PART_MAX; i++) {
part = sc->part[i];
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
if (part != NULL) {
if (part->disk != NULL) {
MMCSD_DISK_LOCK(part);
part->suspend = 1;
if (part->running > 0) {
/* kill thread */
part->running = 0;
wakeup(part);
/* wait for thread to finish. */
while (part->running != -1)
msleep(part, &part->disk_mtx, 0,
"mmcsd disk suspension", 0);
}
MMCSD_DISK_UNLOCK(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
}
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_IOCTL_LOCK(part);
while (part->ioctl > 0)
msleep(part, &part->ioctl_mtx, 0,
"mmcsd IOCTL suspension", 0);
part->ioctl = -1;
MMCSD_IOCTL_UNLOCK(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
}
}
return (0);
}
static int
mmcsd_resume(device_t dev)
{
struct mmcsd_softc *sc = device_get_softc(dev);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
struct mmcsd_part *part;
int i;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
for (i = 0; i < MMC_PART_MAX; i++) {
part = sc->part[i];
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
if (part != NULL) {
if (part->disk != NULL) {
MMCSD_DISK_LOCK(part);
part->suspend = 0;
if (part->running <= 0) {
part->running = 1;
MMCSD_DISK_UNLOCK(part);
kproc_create(&mmcsd_task, part,
&part->p, 0, 0, "%s%d: mmc/sd card",
part->name, part->cnt);
} else
MMCSD_DISK_UNLOCK(part);
}
MMCSD_IOCTL_LOCK(part);
part->ioctl = 0;
MMCSD_IOCTL_UNLOCK(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
}
}
return (0);
}
static int
mmcsd_open(struct disk *dp __unused)
{
return (0);
}
static int
mmcsd_close(struct disk *dp __unused)
{
return (0);
}
static void
mmcsd_strategy(struct bio *bp)
{
struct mmcsd_softc *sc;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
struct mmcsd_part *part;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
part = bp->bio_disk->d_drv1;
sc = part->sc;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_DISK_LOCK(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
if (part->running > 0 || part->suspend > 0) {
bioq_disksort(&part->bio_queue, bp);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_DISK_UNLOCK(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
wakeup(part);
} else {
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_DISK_UNLOCK(part);
biofinish(bp, NULL, ENXIO);
}
}
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
static int
mmcsd_ioctl_rpmb(struct cdev *dev, u_long cmd, caddr_t data,
int fflag, struct thread *td __unused)
{
return (mmcsd_ioctl(dev->si_drv1, cmd, data, fflag));
}
static int
mmcsd_ioctl_disk(struct disk *disk, u_long cmd, void *data, int fflag,
struct thread *td __unused)
{
return (mmcsd_ioctl(disk->d_drv1, cmd, data, fflag));
}
static int
mmcsd_ioctl(struct mmcsd_part *part, u_long cmd, void *data, int fflag)
{
struct mmc_ioc_cmd *mic;
struct mmc_ioc_multi_cmd *mimc;
int i, err;
u_long cnt, size;
if ((fflag & FREAD) == 0)
return (EBADF);
err = 0;
switch (cmd) {
case MMC_IOC_CMD:
mic = data;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
err = mmcsd_ioctl_cmd(part, mic, fflag);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
break;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
case MMC_IOC_MULTI_CMD:
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
mimc = data;
if (mimc->num_of_cmds == 0)
break;
if (mimc->num_of_cmds > MMC_IOC_MAX_CMDS)
return (EINVAL);
cnt = mimc->num_of_cmds;
size = sizeof(*mic) * cnt;
mic = malloc(size, M_TEMP, M_WAITOK);
err = copyin((const void *)mimc->cmds, mic, size);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
if (err == 0) {
for (i = 0; i < cnt; i++) {
err = mmcsd_ioctl_cmd(part, &mic[i], fflag);
if (err != 0)
break;
}
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
}
free(mic, M_TEMP);
break;
default:
return (ENOIOCTL);
}
return (err);
}
static int
mmcsd_ioctl_cmd(struct mmcsd_part *part, struct mmc_ioc_cmd *mic, int fflag)
{
struct mmc_command cmd;
struct mmc_data data;
struct mmcsd_softc *sc;
device_t dev, mmcbr;
void *dp;
u_long len;
int err, retries;
uint32_t status;
uint16_t rca;
if ((fflag & FWRITE) == 0 && mic->write_flag != 0)
return (EBADF);
if (part->ro == TRUE && mic->write_flag != 0)
return (EROFS);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
/*
* We don't need to explicitly lock against the disk(9) half of this
* driver as MMCBUS_ACQUIRE_BUS() will serialize us. However, it's
* necessary to protect against races with detachment and suspension,
* especially since it's required to switch away from RPMB partitions
* again after an access (see mmcsd_switch_part()).
*/
MMCSD_IOCTL_LOCK(part);
while (part->ioctl != 0) {
if (part->ioctl < 0) {
MMCSD_IOCTL_UNLOCK(part);
return (ENXIO);
}
msleep(part, &part->ioctl_mtx, 0, "mmcsd IOCTL", 0);
}
part->ioctl = 1;
MMCSD_IOCTL_UNLOCK(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
err = 0;
dp = NULL;
len = mic->blksz * mic->blocks;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
if (len > MMC_IOC_MAX_BYTES) {
err = EOVERFLOW;
goto out;
}
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
if (len != 0) {
dp = malloc(len, M_TEMP, M_WAITOK);
err = copyin((void *)(uintptr_t)mic->data_ptr, dp, len);
if (err != 0)
goto out;
}
memset(&cmd, 0, sizeof(cmd));
memset(&data, 0, sizeof(data));
cmd.opcode = mic->opcode;
cmd.arg = mic->arg;
cmd.flags = mic->flags;
if (len != 0) {
data.len = len;
data.data = dp;
data.flags = mic->write_flag != 0 ? MMC_DATA_WRITE :
MMC_DATA_READ;
cmd.data = &data;
}
sc = part->sc;
rca = sc->rca;
if (mic->is_acmd == 0) {
/* Enforce/patch/restrict RCA-based commands */
switch (cmd.opcode) {
case MMC_SET_RELATIVE_ADDR:
case MMC_SELECT_CARD:
err = EPERM;
goto out;
case MMC_STOP_TRANSMISSION:
if ((cmd.arg & 0x1) == 0)
break;
/* FALLTHROUGH */
case MMC_SLEEP_AWAKE:
case MMC_SEND_CSD:
case MMC_SEND_CID:
case MMC_SEND_STATUS:
case MMC_GO_INACTIVE_STATE:
case MMC_FAST_IO:
case MMC_APP_CMD:
cmd.arg = (cmd.arg & 0x0000FFFF) | (rca << 16);
break;
default:
break;
}
}
dev = sc->dev;
mmcbr = sc->mmcbr;
MMCBUS_ACQUIRE_BUS(mmcbr, dev);
err = mmcsd_switch_part(mmcbr, dev, rca, part->type);
if (err != MMC_ERR_NONE)
goto release;
if (part->type == EXT_CSD_PART_CONFIG_ACC_RPMB) {
err = mmcsd_set_blockcount(sc, mic->blocks,
mic->write_flag & (1 << 31));
if (err != MMC_ERR_NONE)
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
goto switch_back;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
}
if (mic->is_acmd != 0)
(void)mmc_wait_for_app_cmd(mmcbr, dev, rca, &cmd, 0);
else
(void)mmc_wait_for_cmd(mmcbr, dev, &cmd, 0);
if (part->type == EXT_CSD_PART_CONFIG_ACC_RPMB) {
/*
* If the request went to the RPMB partition, try to ensure
* that the command actually has completed ...
*/
retries = MMCSD_CMD_RETRIES;
do {
err = mmc_send_status(mmcbr, dev, rca, &status);
if (err != MMC_ERR_NONE)
break;
if (R1_STATUS(status) == 0 &&
R1_CURRENT_STATE(status) != R1_STATE_PRG)
break;
DELAY(1000);
} while (retries-- > 0);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
switch_back:
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
/* ... and always switch back to the default partition. */
err = mmcsd_switch_part(mmcbr, dev, rca,
EXT_CSD_PART_CONFIG_ACC_DEFAULT);
if (err != MMC_ERR_NONE)
goto release;
}
/*
* If EXT_CSD was changed, our copy is outdated now. Specifically,
* the upper bits of EXT_CSD_PART_CONFIG used in mmcsd_switch_part(),
* so retrieve EXT_CSD again.
*/
if (cmd.opcode == MMC_SWITCH_FUNC) {
err = mmc_send_ext_csd(mmcbr, dev, sc->ext_csd);
if (err != MMC_ERR_NONE)
goto release;
}
MMCBUS_RELEASE_BUS(mmcbr, dev);
if (cmd.error != MMC_ERR_NONE) {
switch (cmd.error) {
case MMC_ERR_TIMEOUT:
err = ETIMEDOUT;
break;
case MMC_ERR_BADCRC:
err = EILSEQ;
break;
case MMC_ERR_INVALID:
err = EINVAL;
break;
case MMC_ERR_NO_MEMORY:
err = ENOMEM;
break;
default:
err = EIO;
break;
}
goto out;
}
memcpy(mic->response, cmd.resp, 4 * sizeof(uint32_t));
if (mic->write_flag == 0 && len != 0) {
err = copyout(dp, (void *)(uintptr_t)mic->data_ptr, len);
if (err != 0)
goto out;
}
goto out;
release:
MMCBUS_RELEASE_BUS(mmcbr, dev);
err = EIO;
out:
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_IOCTL_LOCK(part);
part->ioctl = 0;
MMCSD_IOCTL_UNLOCK(part);
wakeup(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
if (dp != NULL)
free(dp, M_TEMP);
return (err);
}
static int
mmcsd_getattr(struct bio *bp)
{
struct mmcsd_part *part;
device_t dev;
if (strcmp(bp->bio_attribute, "MMC::device") == 0) {
if (bp->bio_length != sizeof(dev))
return (EFAULT);
part = bp->bio_disk->d_drv1;
dev = part->sc->dev;
bcopy(&dev, bp->bio_data, sizeof(dev));
bp->bio_completed = bp->bio_length;
return (0);
}
return (-1);
}
static int
mmcsd_set_blockcount(struct mmcsd_softc *sc, u_int count, bool reliable)
{
struct mmc_command cmd;
struct mmc_request req;
memset(&req, 0, sizeof(req));
memset(&cmd, 0, sizeof(cmd));
cmd.mrq = &req;
req.cmd = &cmd;
cmd.opcode = MMC_SET_BLOCK_COUNT;
cmd.arg = count & 0x0000FFFF;
if (reliable)
cmd.arg |= 1 << 31;
cmd.flags = MMC_RSP_R1 | MMC_CMD_AC;
MMCBUS_WAIT_FOR_REQUEST(sc->mmcbr, sc->dev, &req);
return (cmd.error);
}
static int
mmcsd_switch_part(device_t bus, device_t dev, uint16_t rca, u_int part)
{
struct mmcsd_softc *sc;
int err;
uint8_t value;
sc = device_get_softc(dev);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
if (sc->mode == mode_sd)
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
return (MMC_ERR_NONE);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
/*
* According to section "6.2.2 Command restrictions" of the eMMC
* specification v5.1, CMD19/CMD21 aren't allowed to be used with
* RPMB partitions. So we pause re-tuning along with triggering
* it up-front to decrease the likelihood of re-tuning becoming
* necessary while accessing an RPMB partition. Consequently, an
* RPMB partition should immediately be switched away from again
* after an access in order to allow for re-tuning to take place
* anew.
*/
if (part == EXT_CSD_PART_CONFIG_ACC_RPMB)
MMCBUS_RETUNE_PAUSE(sc->mmcbr, sc->dev, true);
if (sc->part_curr == part)
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
return (MMC_ERR_NONE);
value = (sc->ext_csd[EXT_CSD_PART_CONFIG] &
~EXT_CSD_PART_CONFIG_ACC_MASK) | part;
/* Jump! */
err = mmc_switch(bus, dev, rca, EXT_CSD_CMD_SET_NORMAL,
EXT_CSD_PART_CONFIG, value, sc->part_time, true);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
if (err != MMC_ERR_NONE) {
if (part == EXT_CSD_PART_CONFIG_ACC_RPMB)
MMCBUS_RETUNE_UNPAUSE(sc->mmcbr, sc->dev);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
return (err);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
}
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
sc->ext_csd[EXT_CSD_PART_CONFIG] = value;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
if (sc->part_curr == EXT_CSD_PART_CONFIG_ACC_RPMB)
MMCBUS_RETUNE_UNPAUSE(sc->mmcbr, sc->dev);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
sc->part_curr = part;
return (MMC_ERR_NONE);
}
static const char *
mmcsd_errmsg(int e)
{
if (e < 0 || e > MMC_ERR_MAX)
return "Bad error code";
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
return (errmsg[e]);
}
static daddr_t
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
mmcsd_rw(struct mmcsd_part *part, struct bio *bp)
{
daddr_t block, end;
struct mmc_command cmd;
struct mmc_command stop;
struct mmc_request req;
struct mmc_data data;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
struct mmcsd_softc *sc;
device_t dev, mmcbr;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
u_int numblocks, sz;
char *vaddr;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
sc = part->sc;
dev = sc->dev;
mmcbr = sc->mmcbr;
block = bp->bio_pblkno;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
sz = part->disk->d_sectorsize;
end = bp->bio_pblkno + (bp->bio_bcount / sz);
while (block < end) {
vaddr = bp->bio_data + (block - bp->bio_pblkno) * sz;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
numblocks = min(end - block, sc->max_data);
memset(&req, 0, sizeof(req));
memset(&cmd, 0, sizeof(cmd));
memset(&stop, 0, sizeof(stop));
memset(&data, 0, sizeof(data));
cmd.mrq = &req;
req.cmd = &cmd;
cmd.data = &data;
if (bp->bio_cmd == BIO_READ) {
if (numblocks > 1)
cmd.opcode = MMC_READ_MULTIPLE_BLOCK;
else
cmd.opcode = MMC_READ_SINGLE_BLOCK;
} else {
if (numblocks > 1)
cmd.opcode = MMC_WRITE_MULTIPLE_BLOCK;
else
cmd.opcode = MMC_WRITE_BLOCK;
}
cmd.arg = block;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
if (sc->high_cap == 0)
cmd.arg <<= 9;
cmd.flags = MMC_RSP_R1 | MMC_CMD_ADTC;
data.data = vaddr;
data.mrq = &req;
if (bp->bio_cmd == BIO_READ)
data.flags = MMC_DATA_READ;
else
data.flags = MMC_DATA_WRITE;
data.len = numblocks * sz;
if (numblocks > 1) {
data.flags |= MMC_DATA_MULTI;
stop.opcode = MMC_STOP_TRANSMISSION;
stop.arg = 0;
stop.flags = MMC_RSP_R1B | MMC_CMD_AC;
stop.mrq = &req;
req.stop = &stop;
}
MMCBUS_WAIT_FOR_REQUEST(mmcbr, dev, &req);
if (req.cmd->error != MMC_ERR_NONE) {
if (ppsratecheck(&sc->log_time, &sc->log_count,
LOG_PPS))
device_printf(dev, "Error indicated: %d %s\n",
req.cmd->error,
mmcsd_errmsg(req.cmd->error));
break;
}
block += numblocks;
}
return (block);
}
static daddr_t
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
mmcsd_delete(struct mmcsd_part *part, struct bio *bp)
{
daddr_t block, end, start, stop;
struct mmc_command cmd;
struct mmc_request req;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
struct mmcsd_softc *sc;
device_t dev, mmcbr;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
u_int erase_sector, sz;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
sc = part->sc;
dev = sc->dev;
mmcbr = sc->mmcbr;
block = bp->bio_pblkno;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
sz = part->disk->d_sectorsize;
end = bp->bio_pblkno + (bp->bio_bcount / sz);
/* Coalesce with part remaining from previous request. */
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
if (block > part->eblock && block <= part->eend)
block = part->eblock;
if (end >= part->eblock && end < part->eend)
end = part->eend;
/* Safe round to the erase sector boundaries. */
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
erase_sector = sc->erase_sector;
start = block + erase_sector - 1; /* Round up. */
start -= start % erase_sector;
stop = end; /* Round down. */
stop -= end % erase_sector;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
/* We can't erase an area smaller than a sector, store it for later. */
if (start >= stop) {
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
part->eblock = block;
part->eend = end;
return (end);
}
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
/*
* Pause re-tuning so it won't interfere with the order of erase
* commands. Note that these latter don't use the data lines, so
* re-tuning shouldn't actually become necessary during erase.
*/
MMCBUS_RETUNE_PAUSE(mmcbr, dev, false);
/* Set erase start position. */
memset(&req, 0, sizeof(req));
memset(&cmd, 0, sizeof(cmd));
cmd.mrq = &req;
req.cmd = &cmd;
if (mmc_get_card_type(dev) == mode_sd)
cmd.opcode = SD_ERASE_WR_BLK_START;
else
cmd.opcode = MMC_ERASE_GROUP_START;
cmd.arg = start;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
if (sc->high_cap == 0)
cmd.arg <<= 9;
cmd.flags = MMC_RSP_R1 | MMC_CMD_AC;
MMCBUS_WAIT_FOR_REQUEST(mmcbr, dev, &req);
if (req.cmd->error != MMC_ERR_NONE) {
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
device_printf(dev, "Setting erase start position failed %d\n",
req.cmd->error);
block = bp->bio_pblkno;
goto unpause;
}
/* Set erase stop position. */
memset(&req, 0, sizeof(req));
memset(&cmd, 0, sizeof(cmd));
req.cmd = &cmd;
if (mmc_get_card_type(dev) == mode_sd)
cmd.opcode = SD_ERASE_WR_BLK_END;
else
cmd.opcode = MMC_ERASE_GROUP_END;
cmd.arg = stop;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
if (sc->high_cap == 0)
cmd.arg <<= 9;
cmd.arg--;
cmd.flags = MMC_RSP_R1 | MMC_CMD_AC;
MMCBUS_WAIT_FOR_REQUEST(mmcbr, dev, &req);
if (req.cmd->error != MMC_ERR_NONE) {
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
device_printf(dev, "Setting erase stop position failed %d\n",
req.cmd->error);
block = bp->bio_pblkno;
goto unpause;
}
/* Erase range. */
memset(&req, 0, sizeof(req));
memset(&cmd, 0, sizeof(cmd));
req.cmd = &cmd;
cmd.opcode = MMC_ERASE;
cmd.arg = 0;
cmd.flags = MMC_RSP_R1B | MMC_CMD_AC;
MMCBUS_WAIT_FOR_REQUEST(mmcbr, dev, &req);
if (req.cmd->error != MMC_ERR_NONE) {
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
device_printf(dev, "erase err3: %d\n", req.cmd->error);
device_printf(dev, "Issuing erase command failed %d\n",
req.cmd->error);
block = bp->bio_pblkno;
goto unpause;
}
/* Store one of remaining parts for the next call. */
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
if (bp->bio_pblkno >= part->eblock || block == start) {
part->eblock = stop; /* Predict next forward. */
part->eend = end;
} else {
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
part->eblock = block; /* Predict next backward. */
part->eend = start;
}
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
block = end;
unpause:
MMCBUS_RETUNE_UNPAUSE(mmcbr, dev);
return (block);
}
static int
mmcsd_dump(void *arg, void *virtual, vm_offset_t physical, off_t offset,
size_t length)
{
struct bio bp;
daddr_t block, end;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
struct disk *disk;
struct mmcsd_softc *sc;
struct mmcsd_part *part;
device_t dev, mmcbr;
int err;
/* length zero is special and really means flush buffers to media */
if (!length)
return (0);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
disk = arg;
part = disk->d_drv1;
sc = part->sc;
dev = sc->dev;
mmcbr = sc->mmcbr;
g_reset_bio(&bp);
bp.bio_disk = disk;
bp.bio_pblkno = offset / disk->d_sectorsize;
bp.bio_bcount = length;
bp.bio_data = virtual;
bp.bio_cmd = BIO_WRITE;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
end = bp.bio_pblkno + bp.bio_bcount / disk->d_sectorsize;
MMCBUS_ACQUIRE_BUS(mmcbr, dev);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
err = mmcsd_switch_part(mmcbr, dev, sc->rca, part->type);
if (err != MMC_ERR_NONE) {
if (ppsratecheck(&sc->log_time, &sc->log_count, LOG_PPS))
device_printf(dev, "Partition switch error\n");
MMCBUS_RELEASE_BUS(mmcbr, dev);
return (EIO);
}
block = mmcsd_rw(part, &bp);
MMCBUS_RELEASE_BUS(mmcbr, dev);
return ((end < block) ? EIO : 0);
}
static void
mmcsd_task(void *arg)
{
daddr_t block, end;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
struct mmcsd_part *part;
struct mmcsd_softc *sc;
struct bio *bp;
device_t dev, mmcbr;
int err, sz;
part = arg;
sc = part->sc;
dev = sc->dev;
mmcbr = sc->mmcbr;
while (1) {
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_DISK_LOCK(part);
do {
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
if (part->running == 0)
goto out;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
bp = bioq_takefirst(&part->bio_queue);
if (bp == NULL)
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
msleep(part, &part->disk_mtx, PRIBIO,
"mmcsd disk jobqueue", 0);
} while (bp == NULL);
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_DISK_UNLOCK(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
if (bp->bio_cmd != BIO_READ && part->ro) {
bp->bio_error = EROFS;
bp->bio_resid = bp->bio_bcount;
bp->bio_flags |= BIO_ERROR;
biodone(bp);
continue;
}
MMCBUS_ACQUIRE_BUS(mmcbr, dev);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
sz = part->disk->d_sectorsize;
block = bp->bio_pblkno;
end = bp->bio_pblkno + (bp->bio_bcount / sz);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
err = mmcsd_switch_part(mmcbr, dev, sc->rca, part->type);
if (err != MMC_ERR_NONE) {
if (ppsratecheck(&sc->log_time, &sc->log_count,
LOG_PPS))
device_printf(dev, "Partition switch error\n");
goto release;
}
if (bp->bio_cmd == BIO_READ || bp->bio_cmd == BIO_WRITE) {
/* Access to the remaining erase block obsoletes it. */
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
if (block < part->eend && end > part->eblock)
part->eblock = part->eend = 0;
block = mmcsd_rw(part, bp);
} else if (bp->bio_cmd == BIO_DELETE) {
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
block = mmcsd_delete(part, bp);
}
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
release:
MMCBUS_RELEASE_BUS(mmcbr, dev);
if (block < end) {
bp->bio_error = EIO;
bp->bio_resid = (end - block) * sz;
bp->bio_flags |= BIO_ERROR;
} else {
bp->bio_resid = 0;
}
biodone(bp);
}
out:
/* tell parent we're done */
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
part->running = -1;
o Add support for eMMC HS200 and HS400 bus speed modes at 200 MHz to sdhci(4), mmc(4) and mmcsd(4). For the most part, this consists of: - Correcting and extending the infrastructure for negotiating and enabling post-DDR52 modes already added as part of r315598. In fact, HS400ES now should work as well but hasn't been activated due to lack of corresponding hardware. - Adding support executing standard SDHCI initial tuning as well as re-tuning as required for eMMC HS200/HS400 and the fast UHS-I SD card modes. Currently, corresponding methods are only hooked up to the ACPI and PCI front-ends of sdhci(4), though. Moreover, sdhci(4) won't offer any modes requiring (re-)tuning to the MMC/SD layer in order to not break operations with other sdhci(4) front- ends. Likewise, sdhci(4) now no longer offers modes requiring the set_uhs_timing method introduced in r315598 to be implemented/ hooked up (previously, this method was used with DDR52 only, which in turn is only available with Intel controllers so far, i. e. no such limitation was necessary before). Similarly for 1.2/1.8 V VCCQ support and the switch_vccq method. - Addition of locking to the IOCTL half of mmcsd(4) to prevent races with detachment and suspension, especially since it's required to immediately switch away from RPMB partitions again after an access to these (so re-tuning can take place anew, given that the current eMMC specification v5.1 doesn't allow tuning commands to be issued with a RPMB partition selected). Therefore, the existing part_mtx lock in the mmcsd(4) softc is additionally renamed to disk_mtx in order to denote that it only refers to the disk(9) half, likewise for corresponding macros. On the system where the addition of DDR52 support increased the read throughput to ~80 MB/s (from ~45 MB/s at high speed), HS200 yields ~154 MB/s and HS400 ~187 MB/s, i. e. performance now has more than quadrupled compared to pre-r315598. Also, with the advent of (re-)tuning support, most infrastructure necessary for SD card UHS-I modes up to SDR104 now is also in place. Note, though, that the standard SDHCI way of (re-)tuning is special in several ways, which also is why sending the actual tuning requests to the device is part of sdhci(4). SDHCI implementations not following the specification, MMC and non-SDHCI SD card controllers likely will use a generic implementation in the MMC/SD layer for executing tuning, which hasn't been written so far, though. However, in fact this isn't a feature-only change; there are boards based on Intel Bay Trail where DDR52 is problematic and the suggested workaround is to use HS200 mode instead. So far exact details are unknown, however, i. e. whether that's due to a defect in these SoCs or on the boards. Moreover, due to the above changes requiring to be aware of possible MMC siblings in the fast path of mmc(4), corresponding information now is cached in mmc_softc. As a side-effect, mmc_calculate_clock(), mmc_delete_cards(), mmc_discover_cards() and mmc_rescan_cards() now all are guaranteed to operate on the same set of devices as there no longer is any use of device_get_children(9), which can fail in low memory situations. Likewise, mmc_calculate_clock() now longer will trigger a panic due to the latter. o Fix a bug in the failure reporting of mmcsd_delete(); in case of an error when the starting block of a previously stored erase request is used (in order to be able to erase a full erase sector worth of data), the starting block of the newly supplied bio_pblkno has to be returned for indicating no progress. Otherwise, upper layers might be told that a negative number of BIOs have been completed, leading to a panic. o Fix 2 bugs on resume: - Things done in fork1(9) like the acquisition of an SX lock or the sleepable memory allocation are incompatible with a MTX_DEF taken. Thus, mmcsd_resume() must not call kproc_create(9), which in turn uses fork1(9), with the disk_mtx (formerly part_mtx) held. - In mmc_suspend(), the bus is powered down, which in the typical case of a device being selected at the time of suspension, causes the device deselection as part of the bus acquisition by mmc(4) in mmc_scan() to fail as the bus isn't powered up again before later in mmc_go_discovery(). Thus, power down with the bus acquired in mmc_suspend(), which will trigger the deselection up-front. o Fix a memory leak in mmcsd_ioctl() in case copyin(9) fails. [1] o Fix missing variable initialization in mmc_switch_status(). [2] o Fix R1_SWITCH_ERROR detection in mmc_switch_status(). [3] o Handle the case of device_add_child(9) failing, for example due to a memory shortage, gracefully in mmc(4) and sdhci(4), including not leaking memory for the instance variables in case of mmc(4) (which might or might not fix [4] as the latter problem has been discovered independently). o Handle the case of an unknown SD CSD version in mmc_decode_csd_sd() gracefully instead of calling panic(9). o Again, check and handle the return values of some additional function calls in mmc(4) instead of assuming that everything went right or mark non-fatal errors by casting the return value to void. o Correct a typo in the Linux IOCTL compatibility; it should have been MMC_IOC_MULTI_CMD rather than MMC_IOC_CMD_MULTI. o Now that we are reaching ever faster speeds (more improvement in this regard is to be expected when adding ADMA support to sdhci(4)), apply a few micro-optimizations like predicting mmc(4) and sdhci(4) debugging to be off or caching erase sector and maximum data sizes as well support of block addressing in mmsd(4) (instead of doing 2 indirections on every read/write request for determining the maximum data size for example). Reported by: Coverity CID: 1372612 [1], 1372624 [2], 1372594 [3], 1007069 [4]
2017-07-23 16:11:47 +00:00
MMCSD_DISK_UNLOCK(part);
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
wakeup(part);
kproc_exit(0);
}
static int
mmcsd_bus_bit_width(device_t dev)
{
if (mmc_get_bus_width(dev) == bus_width_1)
return (1);
if (mmc_get_bus_width(dev) == bus_width_4)
return (4);
return (8);
}
static device_method_t mmcsd_methods[] = {
DEVMETHOD(device_probe, mmcsd_probe),
DEVMETHOD(device_attach, mmcsd_attach),
DEVMETHOD(device_detach, mmcsd_detach),
DEVMETHOD(device_suspend, mmcsd_suspend),
DEVMETHOD(device_resume, mmcsd_resume),
DEVMETHOD_END
};
static driver_t mmcsd_driver = {
"mmcsd",
mmcsd_methods,
sizeof(struct mmcsd_softc),
};
static devclass_t mmcsd_devclass;
- Add support for eMMC "partitions". Besides the user data area, i. e. the default partition, eMMC v4.41 and later devices can additionally provide up to: 1 enhanced user data area partition 2 boot partitions 1 RPMB (Replay Protected Memory Block) partition 4 general purpose partitions (optionally with a enhanced or extended attribute) Of these "partitions", only the enhanced user data area one actually slices the user data area partition and, thus, gets handled with the help of geom_flashmap(4). The other types of partitions have address space independent from the default partition and need to be switched to via CMD6 (SWITCH), i. e. constitute a set of additional "disks". The second kind of these "partitions" doesn't fit that well into the design of mmc(4) and mmcsd(4). I've decided to let mmcsd(4) hook all of these "partitions" up as disk(9)'s (except for the RPMB partition as it didn't seem to make much sense to be able to put a file-system there and may require authentication; therefore, RPMB partitions are solely accessible via the newly added IOCTL interface currently; see also below). This approach for one resulted in cleaner code. Second, it retains the notion of mmcsd(4) children corresponding to a single physical device each. With the addition of some layering violations, it also would have been possible for mmc(4) to add separate mmcsd(4) instances with one disk each for all of these "partitions", however. Still, both mmc(4) and mmcsd(4) share some common code now e. g. for issuing CMD6, which has been factored out into mmc_subr.c. Besides simply subdividing eMMC devices, some Intel NUCs having UEFI code in the boot partitions etc., another use case for the partition support is the activation of pseudo-SLC mode, which manufacturers of eMMC chips typically associate with the enhanced user data area and/ or the enhanced attribute of general purpose partitions. CAVEAT EMPTOR: Partitioning eMMC devices is a one-time operation. - Now that properly issuing CMD6 is crucial (so data isn't written to the wrong partition for example), make a step into the direction of correctly handling the timeout for these commands in the MMC layer. Also, do a SEND_STATUS when CMD6 is invoked with an R1B response as recommended by relevant specifications. However, quite some work is left to be done in this regard; all other R1B-type commands done by the MMC layer also should be followed by a SEND_STATUS (CMD13), the erase timeout calculations/handling as documented in specifications are entirely ignored so far, the MMC layer doesn't provide timeouts applicable up to the bridge drivers and at least sdhci(4) currently is hardcoding 1 s as timeout for all command types unconditionally. Let alone already available return codes often not being checked in the MMC layer ... - Add an IOCTL interface to mmcsd(4); this is sufficiently compatible with Linux so that the GNU mmc-utils can be ported to and used with FreeBSD (note that due to the remaining deficiencies outlined above SANITIZE operations issued by/with `mmc` currently most likely will fail). These latter will be added to ports as sysutils/mmc-utils in a bit. Among others, the `mmc` tool of the GNU mmc-utils allows for partitioning eMMC devices (tested working). - For devices following the eMMC specification v4.41 or later, year 0 is 2013 rather than 1997; so correct this for assembling the device ID string properly. - Let mmcsd.ko depend on mmc.ko. Additionally, bump MMC_VERSION as at least for some of the above a matching pair is required. - In the ACPI front-end of sdhci(4) describe the Intel eMMC and SDXC controllers as such in order to match the PCI one. Additionally, in the entry for the 80860F14 SDXC controller remove the eMMC-only SDHCI_QUIRK_INTEL_POWER_UP_RESET. OKed by: imp Submitted by: ian (mmc_switch_status() implementation)
2017-03-16 22:23:04 +00:00
static int
mmcsd_handler(module_t mod __unused, int what, void *arg __unused)
{
switch (what) {
case MOD_LOAD:
flash_register_slicer(mmcsd_slicer, FLASH_SLICES_TYPE_MMC,
TRUE);
return (0);
case MOD_UNLOAD:
flash_register_slicer(NULL, FLASH_SLICES_TYPE_MMC, TRUE);
return (0);
}
return (0);
}
DRIVER_MODULE(mmcsd, mmc, mmcsd_driver, mmcsd_devclass, mmcsd_handler, NULL);
MODULE_DEPEND(mmcsd, g_flashmap, 0, 0, 0);
MMC_DEPEND(mmcsd);