]> git.ipfire.org Git - thirdparty/u-boot.git/log
thirdparty/u-boot.git
23 months agoclk: mediatek: mt7988: drop 1/1 infracfg spurious factor
Christian Marangi [Sat, 3 Aug 2024 08:32:55 +0000 (10:32 +0200)] 
clk: mediatek: mt7988: drop 1/1 infracfg spurious factor

Now that we can have advanced parent handling for mux, we can drop
spurious infracfg 1/1 factor. This is in preparation to make the clk
ID match the ID in upstream include for mt7988.

Drop the factor entry from mt7988-clk.h and reference to them in
mt7988.dtsi. Muxes and gates are updated to reference the topckgen clk
following how it's done in upstream kernel linux.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7988: fix wrong parent for INFRA_PCIE_PERI_26M_CK_P2
Christian Marangi [Sat, 3 Aug 2024 08:32:54 +0000 (10:32 +0200)] 
clk: mediatek: mt7988: fix wrong parent for INFRA_PCIE_PERI_26M_CK_P2

Fix wrong parent for INFRA_PCIE_PERI_26M_CK_P2 as should be
INFRA_PCIE_PERI_26M_CK_P3 instead of INFRA_F26M_O0. This is to match
implementation on upstream kernel linux.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7988: move INFRA_PCIE_PERI_26M_CK_Px clock at top
Christian Marangi [Sat, 3 Aug 2024 08:32:53 +0000 (10:32 +0200)] 
clk: mediatek: mt7988: move INFRA_PCIE_PERI_26M_CK_Px clock at top

Move INFRA_PCIE_PERI_26M_CK_Px clock at top of the infracfg gates
in preparation for support of OF_UPSTREAM to have a 1:1 match with
upstream clock ID.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7988: rename TOP_CK_NPU_SEL_CM_TOPS_SEL to TOP_NPU_SEL
Christian Marangi [Sat, 3 Aug 2024 08:32:52 +0000 (10:32 +0200)] 
clk: mediatek: mt7988: rename TOP_CK_NPU_SEL_CM_TOPS_SEL to TOP_NPU_SEL

Upstream kernel linux clock include use TOP_NPU_SEL instead of
TOP_CK_NPU_SEL_CM_TOPS_SEL.
Rename this clock to the upstream kernel in preparation for support of
OF_UPSTREAM.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7988: rename TOP_DA_SELM_XTAL_SEL to TOP_DA_SEL
Christian Marangi [Sat, 3 Aug 2024 08:32:51 +0000 (10:32 +0200)] 
clk: mediatek: mt7988: rename TOP_DA_SELM_XTAL_SEL to TOP_DA_SEL

Upstream kernel linux clock include use TOP_DA_SEL instead of
TOP_DA_SELM_XTAL_SEL.
Rename this clock to the upstream kernel in preparation for support of
OF_UPSTREAM.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7988: rename CB_CKSQ_40M to TOP_XTAL
Christian Marangi [Sat, 3 Aug 2024 08:32:50 +0000 (10:32 +0200)] 
clk: mediatek: mt7988: rename CB_CKSQ_40M to TOP_XTAL

Upstream kernel linux clock include use TOP_XTAL instead of CB_CKSQ_40M.
Rename this clock to the upstream kernel in preparation for support of
OF_UPSTREAM.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoMerge patch series "clk: mediatek: mt7981: clk migration for OF_UPSTREAM"
Tom Rini [Mon, 19 Aug 2024 22:13:51 +0000 (16:13 -0600)] 
Merge patch series "clk: mediatek: mt7981: clk migration for OF_UPSTREAM"

Christian Marangi <ansuelsmth@gmail.com> says:

These are all the required patches to migrate clk and correctly support
OF_UPSTREAM. This will align the clk index to upstream to support the same
clk implementation with downstream and upstream DTS.

23 months agoclk: mediatek: mt7981: rename CK to CLK
Christian Marangi [Fri, 2 Aug 2024 13:53:15 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: rename CK to CLK

Rename each entry from CK to CLK to match the include in upstream kernel
linux.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: convert to unified infracfg gates + muxes
Christian Marangi [Fri, 2 Aug 2024 13:53:14 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: convert to unified infracfg gates + muxes

Convert to infracfg gates + muxes implementation now that it's
supported.

Drop infracfg-ao nodes and rename all infracfg-ao clocks to infracfg.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: fix support for pwm3 clock
Christian Marangi [Fri, 2 Aug 2024 13:53:13 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: fix support for pwm3 clock

Add and fix support for pwm3 clock. In the pwm DTSI node we were
actually using PWM2 clock for PWM3. Now that we have correct ID also add
the missing entry of gate and mux to support PWM3 clock.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: replace infracfg ID with upstream linux
Christian Marangi [Fri, 2 Aug 2024 13:53:12 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: replace infracfg ID with upstream linux

Replace infracfg clk ID with upstream linux version.

Add some missing clk for PWM3 and for PCIe. The same format is used here
with the factor first, then mux and then gates.

To correctly reference the gates in clk_gate function, define the
gates_offs value in clk_tree now that they are at an offset from mux and
factor.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: drop 1/1 spurious factor
Christian Marangi [Fri, 2 Aug 2024 13:53:11 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: drop 1/1 spurious factor

Now that we can have advanced parent handling for mux, we can drop
spurious infracfg 1/1 factor. This is in preparation to make the clk
ID match the ID in upstream include for mt7981.

Drop the factor entry from mt7981-clk.h and reference to them in
mt7981.dtsi. Muxes and gates are updated to reference the topckgen clk
following how it's done in upstream kernel linux. Add relevant clk type
flag in clk_tree for infracfg and topckgen.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: implement sgmii0/1 clock
Christian Marangi [Fri, 2 Aug 2024 13:53:10 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: implement sgmii0/1 clock

Implement missing sgmii0/1 clock and update the compatible the DTS to
match upstream kernel linux and in preparation for OF_UPSTREAM support
since the ethernet node define these additional clocks.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: fix wrong parent list for INFRA_PWM1_SEL mux
Christian Marangi [Fri, 2 Aug 2024 13:53:09 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: fix wrong parent list for INFRA_PWM1_SEL mux

Fix wrong parent list for INFRA_PWM1_SEL mux. The list is incorrect and
the parents are just 2. This also match the upstream linux
implementation.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: fix wrong parent for TOP_FAUD clock
Christian Marangi [Fri, 2 Aug 2024 13:53:08 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: fix wrong parent for TOP_FAUD clock

Fix wrong parent for TOP_FAUD clock. Upstream linux sets the parent for
TOP_FAUD to TOP_AUD_SEL instead of CB_CKSQ_40M.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: swap wrong clock-names for spi nodes
Christian Marangi [Fri, 2 Aug 2024 13:53:07 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: swap wrong clock-names for spi nodes

Swap wrong clock-names for spi nodes as they were wrong and the spi-clk
was referencing the sel-clk and the sel-clk was referencing the spi-clk.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: add missing clock for spi1 node
Christian Marangi [Fri, 2 Aug 2024 13:53:06 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: add missing clock for spi1 node

Add missing clock for spi1 node. Clocks taken from upstream clock
definition and clock mux.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: fix wrong clock definition for spi2
Christian Marangi [Fri, 2 Aug 2024 13:53:05 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: fix wrong clock definition for spi2

Fix wrong clock definition for spi2 node as they reference spi0 clock
instead of spi2.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: fix wrong mux width for pwm2 and pwm1 clock
Christian Marangi [Fri, 2 Aug 2024 13:53:04 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: fix wrong mux width for pwm2 and pwm1 clock

Fix wrong mux width for pwm2 and pwm1. Upstream have width 1 but U-Boot
have width set to 2. Change the value to follow upstream implementation.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: fix typo for infra_i2c0_ck
Christian Marangi [Fri, 2 Aug 2024 13:53:03 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: fix typo for infra_i2c0_ck

Fix a typo for infra_i2c0_ck where 0 was misspelled as O.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7981: add missing clock for infra_ipcie_pipe
Christian Marangi [Fri, 2 Aug 2024 13:53:02 +0000 (15:53 +0200)] 
clk: mediatek: mt7981: add missing clock for infra_ipcie_pipe

Add missing clock for infra_ipcie_pipe to make PCIe correctly work. This
clock is a parent of the fixed clock from topckgen cb_cksq_40m.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoMerge patch series "clk: mediatek: mt7623: clk migration for OF_UPSTREAM"
Tom Rini [Mon, 19 Aug 2024 22:13:01 +0000 (16:13 -0600)] 
Merge patch series "clk: mediatek: mt7623: clk migration for OF_UPSTREAM"

Christian Marangi <ansuelsmth@gmail.com> says:

These are all the required patches to migrate clk and correctly support
OF_UPSTREAM. This will align the clk index to upstream to support the same
clk implementation with downstream and upstream DTS.

23 months agoclk: mediatek: mt7623: remap peri clock ID and add MUX
Christian Marangi [Fri, 2 Aug 2024 13:45:05 +0000 (15:45 +0200)] 
clk: mediatek: mt7623: remap peri clock ID and add MUX

Upstream kernel linux makes use of peri clock MUX to setup UART.

Add definition for these and add remap table as in the upstream clock ID
order gates are before MUX but we require MUX first and then clocks in
our downstream driver.

Convert the peri clk tree to MUX + GATE implementation.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Tested-by: Frank Wunderlich <frank-w@public-files.de>
23 months agoclk: mediatek: mt7623: remap apmixedsys clock ID
Christian Marangi [Fri, 2 Aug 2024 13:45:04 +0000 (15:45 +0200)] 
clk: mediatek: mt7623: remap apmixedsys clock ID

Define remap table also for apmixedsys clock ID. The clock ID starts
from 1 instead of 0 in upstream kernel linux.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7623: define id_offs_map and import clk ID from upstream
Christian Marangi [Fri, 2 Aug 2024 13:45:03 +0000 (15:45 +0200)] 
clk: mediatek: mt7623: define id_offs_map and import clk ID from upstream

Define id_offs_map and use clk ID form upstream linux kernel to have a
1:1 match for the TOPCKGEN clock and permit usage of OF_UPSTREAM with
upstream dtsi.

For all the gate clock, the clk ID starts from 1 instead of zero. Define
an additional clock tree for them and set the .gates_offs to 1 to
account for this.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7623: split clk tree to dedicated topckgen and apmixed
Christian Marangi [Fri, 2 Aug 2024 13:45:02 +0000 (15:45 +0200)] 
clk: mediatek: mt7623: split clk tree to dedicated topckgen and apmixed

Split clk tree to dedicated topckgen and apmixed in preparation for
remap table.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agoclk: mediatek: mt7623: fix broken peri_cgs clk with XTAL parents
Christian Marangi [Fri, 2 Aug 2024 13:45:01 +0000 (15:45 +0200)] 
clk: mediatek: mt7623: fix broken peri_cgs clk with XTAL parents

Fix broken peri_cgs and infra_cgs clock with XTAL parents as they have
wrong definition of the parent type.

Correctly fix them and use CLK_PARENT_XTAL for them.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
23 months agodoc: samsung: Mention enabled eMMC in E850-96 board doc
Sam Protsenko [Thu, 8 Aug 2024 03:14:44 +0000 (22:14 -0500)] 
doc: samsung: Mention enabled eMMC in E850-96 board doc

eMMC is enabled on E850-96 board now. Mention that in the board
documentation.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agoconfigs: e850-96: Enable MMC
Sam Protsenko [Thu, 8 Aug 2024 03:14:43 +0000 (22:14 -0500)] 
configs: e850-96: Enable MMC

Enable MMC subsystem and DW MMC driver support to make eMMC functional.
Also enable a couple of related commands so the user can make use of
eMMC from U-Boot shell.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agoarm: dts: exynos: Remove outdated DW MMC properties in all Exynos dts
Sam Protsenko [Thu, 8 Aug 2024 03:14:42 +0000 (22:14 -0500)] 
arm: dts: exynos: Remove outdated DW MMC properties in all Exynos dts

Upstream properties were added to device trees to follow current Linux
kernel. DW MMC driver was updated accordingly. Safely remove outdated
MMC properties. Details on removed properties are as follows:

* samsung,removable: replaced by non-removable
* samsung,bus-width: replaced by bus-width
* samsung,timing:
  - replaced by samsung,dw-mshc-ciu-div and samsung,dw-mshc-sdr-timing
    in dw_mmc nodes
  - removed from sdhci nodes (it's neither described in bindings, nor
    it's used in s5p_sdhci.c driver)
* fifoth_val: replaced by fifo-depth
* bus_hz: replaced by clock-frequency
* div: the fixed CIU clock divider value was moved to the chip data in
  exynos_dw_mmc.c driver

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Improve coding style
Sam Protsenko [Thu, 8 Aug 2024 03:14:41 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Improve coding style

Fix most of checkpatch warnings and other obvious style issues.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Use dev->name as driver's displayed name
Sam Protsenko [Thu, 8 Aug 2024 03:14:40 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Use dev->name as driver's displayed name

Reduce U-Boot footprint by reusing dev->name as a driver's displayed
name. This changes boot device name (and "mmc info" output) from "EXYNOS
DWMMC" to something like "mmc@12100000".

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Don't call dwmci_setup_cfg() after add_dwmci()
Sam Protsenko [Thu, 8 Aug 2024 03:14:39 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Don't call dwmci_setup_cfg() after add_dwmci()

add_dwmci() is already calling dwmci_setup_cfg() internally, there is no
needed to call dwmci_setup_cfg() again in case when add_dwmci() is used
(for non-DM cases). Fix it by calling dwmci_setup_cfg() only in DM
cases, when add_dwmci() wasn't called. Also, this assignment:

    host->mmc = &plat->mmc;

is wrong in non-DM case when add_dwmci() was called, as it's creating
mmc object internally. Fix that by pulling that assignment into DM case,
when add_dwmci() isn't called.

While at it, add also this missing assignment:

    host->mmc->dev = dev;

Fixes: 3537ee879e04 ("mmc: exynos_dw_mmc: support the Driver mode for Exynos")
Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Pull all init code into probe function
Sam Protsenko [Thu, 8 Aug 2024 03:14:38 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Pull all init code into probe function

There is no logical sense to split the initialization code between
multiple functions. Pull both do_dwmci_init() and
exynos_dwmci_core_init() into exynos_dwmmc_probe() to make the code more
simple and obvious.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Add support for ARM64 Exynos chips
Sam Protsenko [Thu, 8 Aug 2024 03:14:37 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Add support for ARM64 Exynos chips

Add the compatible entry and corresponding chip data for Exynos7
compatible chips, which covers modern ARM64 based Exynos chips. They
have some differences w.r.t. old ARM32 Exynos chips:
  - CLKSEL register offset is different
  - 64-bit IDMAC descriptor and 64-bit IDMAC registers are used
    (implemented in dw_mmc core driver)

In terms of the driver implementation, the CIU clock is obtained via CCF
framework (as opposed to ad-hoc clock driver implementation for ARM32
chips).

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Set requested freq in get_mmc_clk() callback
Sam Protsenko [Thu, 8 Aug 2024 03:14:36 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Set requested freq in get_mmc_clk() callback

By now exynos_dw_mmc driver was relying on the correct CIU clock
frequency being set on driver init. But dw_mmc core is actually trying
to change CIU clock rate dynamically, on init and in set_ios() callback,
which it's requesting via host->get_mmc_clk() callback (the name is
misleading: although it's called "get_mmc_clk()", it can actually
request both get and set operations). Implement setting the requested
rate for CIU clock in Exynos driver to achieve the correct dw_mmc core
driver operation at all times. DDR mode requires the clock to be twice
as fast (when 8 bit bus is used), so handle this too, to make DDR
function properly.

This change makes the eMMC throughput on E850-96 board twice as fast.
That's because "clock-frequency" is set to 800 MHz in E850-96 device
tree, but for DDR52 mode it should be 416 MHz (and TRM states it should
be 400 MHz for DDR50/8bit mode). The dw_mmc core is requesting 52 MHz
bus_hz for DDR52 mode, and DDR+8bit mode means it should be x2 fast, so:

    f_ciu = 2 * ciu_div * f_bus = 2 * 4 * 52e6 = 416 MHz,

where f_ciu   - freq of clock fed to DW MMC block from CMU (SDCLKIN), Hz
      f_bus   - freq of clock fed to the card (CCLKIN), Hz
      ciu_div - value of internal divider (in DW MMC block).

Another way to work that around would be overriding the
"clock-frequency" property in corresponding dts. But setting the clock
frequency dynamically as it's done here looks much neater.

This implementation follows what's done in Linux kernel dw_mmc-exynos
driver in .set_ios() callback for MMC_TIMING_MMC_DDR52 case.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Read and use DDR timing when available
Sam Protsenko [Thu, 8 Aug 2024 03:14:35 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Read and use DDR timing when available

DDR timing values should be defined in "samsung,dw-mshc-ddr-timing" dts
property, and used when DDR MMC mode is selected. Read that value from
dts and use it. If it's not available, use SDR timing values instead.
This change is following upstream Linux kernel implementation.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Move quirks from struct dwmci_host to chip data
Sam Protsenko [Thu, 8 Aug 2024 03:14:34 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Move quirks from struct dwmci_host to chip data

host->quirks field is only used internally in exynos_dw_mmc.c driver.
To avoid cluttering the scope of struct dwmci_host, move quirks field
into Exynos driver's chip data, where it can be statically defined.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Read common clock-frequency property
Sam Protsenko [Thu, 8 Aug 2024 03:14:33 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Read common clock-frequency property

Instead of using non-standard "bus_hz" dts property, read common
"clock-frequency" property used in upstream Linux kernel. It's safe to
do so, as "clock-frequency" property was already added to corresponding
nodes in all affected Exynos device tree files.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Read common bus-width property
Sam Protsenko [Thu, 8 Aug 2024 03:14:32 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Read common bus-width property

Instead of using non-standard "samsung,bus-width" dts property, read
common "bus-width" property used in upstream Linux kernel. It's safe to
do so, as "bus-width" property was already added to corresponding nodes
in all affected Exynos device tree files.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Refactor fixed CIU clock divider
Sam Protsenko [Thu, 8 Aug 2024 03:14:31 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Refactor fixed CIU clock divider

Some chips like Exynos4412 have fixed internal CIU clock divider.
Instead of reading it from non-standard "div" dts property, store its
value in the driver internally, in static chip data associated with
corresponding compatible. This makes it possible to avoid using
host->div for storing it, so the latter can be removed safely. Also
create a helper function called exynos_dwmmc_get_ciu_div() for getting
the current div value: in case the fixed div is provided in the chip
data it will be used, otherwise the current div value is being read from
CLKSEL register.

The insights for this change were taken from dw_mmc-exynos.c driver in
Linux kernel.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Abstract CLKSEL register
Sam Protsenko [Thu, 8 Aug 2024 03:14:30 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Abstract CLKSEL register

CLKSEL register offset may vary between different Exynos chips, e.g. on
ARM64 vs ARM32 chips. Provide a way to specify its offset value for each
compatible instead of hard-coding its value in read/write calls.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Read upstream SDR timing properties
Sam Protsenko [Thu, 8 Aug 2024 03:14:29 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Read upstream SDR timing properties

The obsolete "samsung,timing" dts property is now split into
"samsung,dw-mshc-ciu-div" (for holding the internal DW MMC divider
value) and "samsung,dw-mshc-sdr-timing" (for actual timing values) in
upstream Linux kernel. Rework the driver to make use of new properties
instead of the old one. All affected dts files were already updated
accordingly.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Convert to use livetree API
Sam Protsenko [Thu, 8 Aug 2024 03:14:28 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Convert to use livetree API

Update the driver to use livetree API instead of FDT one.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Use .of_to_plat for device tree parsing
Sam Protsenko [Thu, 8 Aug 2024 03:14:27 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Use .of_to_plat for device tree parsing

exynos_dwmci_get_config() is called from the probe function and used to
read data from device tree. Make use of .of_to_plat driver callback
instead, and convert exynos_dwmci_get_config() to match its signature.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Obtain and use CIU clock via CCF API
Sam Protsenko [Thu, 8 Aug 2024 03:14:26 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Obtain and use CIU clock via CCF API

New Exynos chips should implement clock drivers using CCF framework. In
that case corresponding CCF functions can be used to get/set the clock
rates. Moreover, already existing get_mmc_clk() and set_mmc_clk() calls
are only implemented for CONFIG_CPU_V7A (i.e. ARM32 chips). In case of
ARM64 chips that config option is not defined, so build will crash on
linking stage, with errors like these:

    ld: drivers/mmc/exynos_dw_mmc.o:
      in function `exynos_dwmci_get_sclk':
      undefined reference to `get_mmc_clk'
    ld: drivers/mmc/exynos_dw_mmc.o:
      in function `exynos_dwmci_set_sclk':
      undefined reference to `set_mmc_clk'

Fix that issue by using CCF clocks API on ARM64 platforms for getting
and setting the source clock (sclk = SDCLKIN = CIU) rate. To implement
this, first extract the existing ARM32 clock control code into helper
functions with more generic signatures to abstract getting/setting the
sclk rate. Then add CCF clock support to those functions for ARM64
platforms.

Fixes: a082a2dde061 ("EXYNOS5: DWMMC: Added FDT support for DWMMC")
Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Don't call pinmux functions on ARM64 chips
Sam Protsenko [Thu, 8 Aug 2024 03:14:25 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Don't call pinmux functions on ARM64 chips

Pinmux configuration on ARM64 platforms must be performed during startup
in pinctrl driver using info from device tree. exynos_pinmux_config()
and pinmux_decode_periph_id() are only available on ARM32 platforms, so
don't call those functions on ARM64 platforms. Instead of the latter
function, use "non-removable" property from device tree to derive the
dev_index value.

This fixes next linking errors on ARM64 platforms:

    ld: drivers/mmc/exynos_dw_mmc.o:
      in function `exynos_dwmci_get_config':
      undefined reference to `pinmux_decode_periph_id'
    ld: drivers/mmc/exynos_dw_mmc.o:
      in function `do_dwmci_init':
      undefined reference to `exynos_pinmux_config'

Fixes: a082a2dde061 ("EXYNOS5: DWMMC: Added FDT support for DWMMC")
Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Fix getting private data in exynos_dwmci_board_init()
Sam Protsenko [Thu, 8 Aug 2024 03:14:24 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Fix getting private data in exynos_dwmci_board_init()

In case of CONFIG_DM_MMC, host->priv actually holds (struct udevice *),
and not (struct dwmci_exynos_priv_data *). This makes *priv pointer
invalid and may lead to Synchronous Abort during its dereference later
in exynos_dwmci_board_init(). Fix it by extracting
exynos_dwmmc_get_priv() helper from exynos_dwmci_clksel() and using it
for getting the private data in exynos_dwmci_board_init()

Fixes: 3537ee879e04 ("mmc: exynos_dw_mmc: support the Driver mode for Exynos")
Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: exynos_dw_mmc: Fix obtaining the base address of controller
Sam Protsenko [Thu, 8 Aug 2024 03:14:23 +0000 (22:14 -0500)] 
mmc: exynos_dw_mmc: Fix obtaining the base address of controller

Getting the base address with outdated fdtdec_get_addr() API and further
casting it to (void *) leads to next build warning on ARM64 platforms:

    In function 'exynos_dwmci_get_config':
        warning: cast to pointer from integer of different size
        [-Wint-to-pointer-cast]
            host->ioaddr = (void *)base;

Use livetree API instead (dev_read_addr_ptr()), which handles this
correctly.

Fixes: a082a2dde061 ("EXYNOS5: DWMMC: Added FDT support for DWMMC")
Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agoarm: exynos: Add header guard for dwmmc.h
Sam Protsenko [Thu, 8 Aug 2024 03:14:22 +0000 (22:14 -0500)] 
arm: exynos: Add header guard for dwmmc.h

Add missing header guard to prevent possible build errors.

Fixes: 77b55e8cfcee ("ARM: exynos: move SoC sources to mach-exynos")
Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agodt-bindings: exynos: Update bindings doc for DW MMC controller
Sam Protsenko [Thu, 8 Aug 2024 03:14:21 +0000 (22:14 -0500)] 
dt-bindings: exynos: Update bindings doc for DW MMC controller

Update the bindings doc for Exynos DW MMC block to follow the upstream
example and reflect the latest changes made in corresponding Linux
kernel bindings.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agoarm: dts: exynos: Add upstream DW MMC properties to all Exynos dts
Sam Protsenko [Thu, 8 Aug 2024 03:14:20 +0000 (22:14 -0500)] 
arm: dts: exynos: Add upstream DW MMC properties to all Exynos dts

Some device tree properties for DW MMC block were updated in Linux
kernel. Let's follow its example and rework corresponding properties in
all Exynos device trees. Don't remove outdated properties yet, it'll be
done later once DW MMC driver is updated accordingly to read the updated
properties instead of outdated ones.

Next properties are added:

* samsung,dw-mshc-ciu-div and samsung,dw-mshc-sdr-timing:

  They were derived from outdated samsung,timing property.

* fifo-depth (generic replacement for fifoth_val):

  FIFO depth was calculated from fifoth_val (using expressions from
  FIFOTH register description in TRM):

      fifo-depth = ((fifoth_val >> 16) + 1) * 2

* bus-width: generic replacement for samsung,bus-width
* clock-frequency: generic replacement for bus_hz
* non-removable: generic replacement for samsung,removable = <0>

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Improve coding style
Sam Protsenko [Thu, 8 Aug 2024 03:14:19 +0000 (22:14 -0500)] 
mmc: dw_mmc: Improve coding style

Fix most of checkpatch warnings and other obvious style issues.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Reviewed-by: Quentin Schulz <quentin.schulz@cherry.de>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Fix kernel-doc comments in dwmmc.h
Sam Protsenko [Thu, 8 Aug 2024 03:14:18 +0000 (22:14 -0500)] 
mmc: dw_mmc: Fix kernel-doc comments in dwmmc.h

Rework kernel-doc comments in dwmmc.h header so it's actually possible
to generate a proper documentation from it usin scripts/kernel-doc
script, with no errors.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Replace fifoth_val property with fifo-depth
Sam Protsenko [Thu, 8 Aug 2024 03:14:17 +0000 (22:14 -0500)] 
mmc: dw_mmc: Replace fifoth_val property with fifo-depth

Replace fifoth_val property with its fifo-depth counterpart in all DW
MMC drivers. fifo-depth is a common property used in upstream Linux
kernel. The FIFOTH register value will be calculated using fifo-depth
value in DW MMC core (dw_mmc.c). This change reduces code duplication in
platform drivers, and pulls common FIFOTH register value calculation
into core dw_mmc driver where it belongs.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Add support for 64-bit IDMAC
Sam Protsenko [Thu, 8 Aug 2024 03:14:16 +0000 (22:14 -0500)] 
mmc: dw_mmc: Add support for 64-bit IDMAC

Some DW MMC blocks (e.g. those on modern Exynos chips) support 64-bit
DMA addressing mode. 64-bit DW MMC variants differ from their 32-bit
counterparts:
  - the register layout is a bit different (because there are additional
    IDMAC registers present for storing upper part of 64-bit addresses)
  - DMA descriptor structure is bigger and different from 32-bit one

Introduce all necessary changes to enable support for 64-bit DMA capable
DW MMC blocks. Next changes were made:

  1. Check which DMA address mode is supported in current IP-core
     version. HCON register (bit 27) indicates whether it's 32-bit or
     64-bit addressing. Add boolean .dma_64bit_address field to struct
     dwmci_host and store the result there. dwmci_init_dma() function is
     introduced for doing so, which is called on driver's init.

  2. Add 64-bit DMA descriptor (struct dwmci_idmac64) and use it in
     dwmci_prepare_desc() in case if .dma_64bit_address field is true.
     A new dwmci_set_idma_desc64() function was added for populating that
     descriptor.

  3. Add registers for 64-bit DMA capable blocks. To make the access to
     IDMAC registers universal between 32-bit / 64-bit cases, a new
     struct dwmci_idmac_regs (and corresponding host->regs field) was
     introduced, which abstracts the hardware by being set to
     appropriate offset constants on init. All direct calls to IDMAC
     registers were correspondingly replaced by accessing host->regs.

  4. Allocate and use 64-bit DMA descriptors buffer in case when IDMAC
     is 64-bit capable. Extract all the code (except for the IDMAC
     descriptors buffer allocation) from dwmci_send_cmd() to
     dwmci_send_cmd_common(), so that it's possible to keep IDMAC
     buffer (either 32-bit or 64-bit) on stack during send_cmd routine.

The insights for this implementation were taken from Linux kernel DW MMC
driver.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Improve 32-bit IDMAC descriptor namings
Sam Protsenko [Thu, 8 Aug 2024 03:14:15 +0000 (22:14 -0500)] 
mmc: dw_mmc: Improve 32-bit IDMAC descriptor namings

Prepare for adding 64-bit IDMAC descriptors by renaming current 32-bit
descriptor and its fields accordingly. While at it, make use of
virt_to_phys() to make it more obvious in which places the physical
addresses have to be used.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Extract setting the DMA descriptor into a separate routine
Sam Protsenko [Thu, 8 Aug 2024 03:14:14 +0000 (22:14 -0500)] 
mmc: dw_mmc: Extract setting the DMA descriptor into a separate routine

Make dwmci_prepare_data() function easier to read by extracting the
preparation of IDMAC descriptor into a dedicated function.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Reviewed-by: Quentin Schulz <quentin.schulz@cherry.de>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Extract DMA transfer handling code into a separate routine
Sam Protsenko [Thu, 8 Aug 2024 03:14:13 +0000 (22:14 -0500)] 
mmc: dw_mmc: Extract DMA transfer handling code into a separate routine

Make dwmci_send_cmd() easier to read by moving the DMA transfer handling
code into a dedicated function.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Reviewed-by: Quentin Schulz <quentin.schulz@cherry.de>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Extract FIFO data transfer into a separate routine
Sam Protsenko [Thu, 8 Aug 2024 03:14:12 +0000 (22:14 -0500)] 
mmc: dw_mmc: Extract FIFO data transfer into a separate routine

FIFO data transfer is implemented as quite a massive chunk of code.
Extract it into a dedicated function to make dwmci_data_transfer()
easier to read and reduce the indentation level of the code.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Extract divider update to a separate function
Sam Protsenko [Thu, 8 Aug 2024 03:14:11 +0000 (22:14 -0500)] 
mmc: dw_mmc: Extract divider update to a separate function

Extract the clock divider update into dwmci_update_div() function. It's
a procedure recommended in TRM, so it's better to keep it in a dedicated
function to make the code clearer.

While at it also extract the clock control code into a separate routine
to avoid code duplication in dwmci_setup_bus().

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Extract FIFO init into a separate routine
Sam Protsenko [Thu, 8 Aug 2024 03:14:10 +0000 (22:14 -0500)] 
mmc: dw_mmc: Extract FIFO init into a separate routine

Move FIFO threshold initialization into a separate function to make
dwmci_init() more readable.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Reviewed-by: Quentin Schulz <quentin.schulz@cherry.de>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Extract waiting for data busy into a separate routine
Sam Protsenko [Thu, 8 Aug 2024 03:14:09 +0000 (22:14 -0500)] 
mmc: dw_mmc: Extract waiting for data busy into a separate routine

Waiting for data busy is a logically separate operation and should be
implemented as a separate routine. Follow Linux kernel example and
extract it from dwmci_send_cmd(). This way it doesn't clutter
dwmci_send_cmd() function, and can be reused later in other cases.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Reviewed-by: Quentin Schulz <quentin.schulz@cherry.de>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Move struct idmac to dw_mmc.c
Sam Protsenko [Thu, 8 Aug 2024 03:14:08 +0000 (22:14 -0500)] 
mmc: dw_mmc: Move struct idmac to dw_mmc.c

struct idmac is only used in dw_mmc.c, so move it there from dwmmc.h to
avoid cluttering the interface in the header.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Reviewed-by: Quentin Schulz <quentin.schulz@cherry.de>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agommc: dw_mmc: Remove unused version field from struct dwmci_host
Sam Protsenko [Thu, 8 Aug 2024 03:14:07 +0000 (22:14 -0500)] 
mmc: dw_mmc: Remove unused version field from struct dwmci_host

Nobody seems to use it, so just remove it.

No functional change.

Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Reviewed-by: Quentin Schulz <quentin.schulz@cherry.de>
Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
23 months agoext4: Fix zalloc()
Richard Weinberger [Fri, 9 Aug 2024 09:54:30 +0000 (11:54 +0200)] 
ext4: Fix zalloc()

Currently, zalloc() calls uncondtionally memset(),
if the allocation failes, memset() will write to a null pointer.

Fix by using kzalloc().

Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
23 months agocompiler: Ensure __builtin_*_overflow() support
Richard Weinberger [Fri, 9 Aug 2024 09:54:29 +0000 (11:54 +0200)] 
compiler: Ensure __builtin_*_overflow() support

Both gcc and clang support this for a long time.
Make sure the feature is present.

Signed-off-by: Richard Weinberger <richard@nod.at>
23 months agoext4: Fix integer overflow in ext4fs_read_symlink()
Richard Weinberger [Fri, 9 Aug 2024 09:54:28 +0000 (11:54 +0200)] 
ext4: Fix integer overflow in ext4fs_read_symlink()

While zalloc() takes a size_t type, adding 1 to the le32 variable
will overflow.
A carefully crafted ext4 filesystem can exhibit an inode size of 0xffffffff
and as consequence zalloc() will do a zero allocation.

Later in the function the inode size is again used for copying data.
So an attacker can overwrite memory.

Avoid the overflow by using the __builtin_add_overflow() helper.

Signed-off-by: Richard Weinberger <richard@nod.at>
23 months agosquashfs: Fix heap corruption in sqfs_search_dir()
Richard Weinberger [Fri, 2 Aug 2024 20:05:09 +0000 (22:05 +0200)] 
squashfs: Fix heap corruption in sqfs_search_dir()

res needs to be large enough to store both strings rem and target,
plus the path separator and the terminator.
Currently the space for the path separator is not accounted, so
the heap is corrupted by one byte.

Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Miquel Raynal <miquel.raynal@bootlin.com>
23 months agosquashfs: Fix stack overflow while symlink resolving
Richard Weinberger [Fri, 2 Aug 2024 16:36:47 +0000 (18:36 +0200)] 
squashfs: Fix stack overflow while symlink resolving

The squashfs driver blindly follows symlinks, and calls sqfs_size()
recursively. So an attacker can create a crafted filesystem and with
a deep enough nesting level a stack overflow can be achieved.

Fix by limiting the nesting level to 8.

Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Miquel Raynal <miquel.raynal@bootlin.com>
23 months agosquashfs: Check sqfs_find_inode() return value
Richard Weinberger [Fri, 2 Aug 2024 16:36:46 +0000 (18:36 +0200)] 
squashfs: Check sqfs_find_inode() return value

The function can fail and return NULL.

Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Miquel Raynal <miquel.raynal@bootlin.com>
23 months agosquashfs: Fix integer overflow in sqfs_inode_size()
Richard Weinberger [Fri, 2 Aug 2024 16:36:45 +0000 (18:36 +0200)] 
squashfs: Fix integer overflow in sqfs_inode_size()

A carefully crafted squashfs filesystem can exhibit an extremly large
inode size and overflow the calculation in sqfs_inode_size().
As a consequence, the squashfs driver will read from wrong locations.

Fix by using __builtin_add_overflow() to detect the overflow.

Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Miquel Raynal <miquel.raynal@bootlin.com>
23 months agosquashfs: Fix integer overflow in sqfs_resolve_symlink()
Richard Weinberger [Fri, 2 Aug 2024 16:36:44 +0000 (18:36 +0200)] 
squashfs: Fix integer overflow in sqfs_resolve_symlink()

A carefully crafted squashfs filesystem can exhibit an inode size of 0xffffffff,
as a consequence malloc() will do a zero allocation.
Later in the function the inode size is again used for copying data.
So an attacker can overwrite memory.
Avoid the overflow by using the __builtin_add_overflow() helper.

Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Miquel Raynal <miquel.raynal@bootlin.com>
23 months agodlmalloc: Make sure allocation size is within malloc area
Richard Weinberger [Fri, 2 Aug 2024 10:08:46 +0000 (12:08 +0200)] 
dlmalloc: Make sure allocation size is within malloc area

Since U-Boot does not support memory overcommit we can
enforce that the allocation size is within the malloc area.
This is a simple and efficient hardening measure to mitigate
further integer overflows in dlmalloc.

Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Simon Glass <sjg@chromium.org>
23 months agodlmalloc: Fix integer overflow in sbrk()
Richard Weinberger [Fri, 2 Aug 2024 10:08:45 +0000 (12:08 +0200)] 
dlmalloc: Fix integer overflow in sbrk()

Make sure that the new break is within mem_malloc_start
and mem_malloc_end before making progress.
ulong new = old + increment; can overflow for extremely large
increment values and memset() can get wrongly called.

Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Simon Glass <sjg@chromium.org>
23 months agodlmalloc: Fix integer overflow in request2size()
Richard Weinberger [Fri, 2 Aug 2024 10:08:44 +0000 (12:08 +0200)] 
dlmalloc: Fix integer overflow in request2size()

req is of type size_t, casting it to long opens the door
for an integer overflow.
Values between LONG_MAX - (SIZE_SZ + MALLOC_ALIGN_MASK) - 1 and LONG_MAX
cause and overflow such that request2size() returns MINSIZE.

Fix by removing the cast.
The origin of the cast is unclear, it's in u-boot and ppcboot since ever
and predates the CVS history.
Doug Lea's original dlmalloc implementation also doesn't have it.

Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Simon Glass <sjg@chromium.org>
23 months agox86: Fix ptrdiff_t for x86_64
Richard Weinberger [Fri, 2 Aug 2024 10:08:43 +0000 (12:08 +0200)] 
x86: Fix ptrdiff_t for x86_64

sbrk() assumes ptrdiff_t is large enough to enlarge/shrink the heap
by LONG_MIN/LONG_MAX.
So, use the long type, also to match the rest of the Linux ecosystem.

Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Simon Glass <sjg@chromium.org>
23 months agoam335x_hs_evm_spi_defconfig: Add MAINTAINERS entry
Tom Rini [Thu, 15 Aug 2024 22:07:45 +0000 (16:07 -0600)] 
am335x_hs_evm_spi_defconfig: Add MAINTAINERS entry

Add this to the existing entry for similar boards.

Signed-off-by: Tom Rini <trini@konsulko.com>
23 months agobootstage: Fix out-of-bounds read in reloc_bootstage()
Richard Weinberger [Wed, 31 Jul 2024 16:07:54 +0000 (18:07 +0200)] 
bootstage: Fix out-of-bounds read in reloc_bootstage()

bootstage_get_size() returns the total size of the data structure
including associated records.
When copying from gd->bootstage, only the allocation size of gd->bootstage
must be used. Otherwise too much memory is copied.

This bug caused no harm so far because gd->new_bootstage is always
large enough and reading beyond the allocation length of gd->bootstage
caused no problem due to the U-Boot memory layout.

Fix by using the correct size and perform the initial copy directly
in bootstage_relocate() to have the whole relocation process in the
same function.

Signed-off-by: Richard Weinberger <richard@nod.at>
Reviewed-by: Simon Glass <sjg@chromium.org>
23 months agoext4: Improve feature checking
Richard Weinberger [Mon, 29 Jul 2024 21:37:31 +0000 (23:37 +0200)] 
ext4: Improve feature checking

Evaluate the filesystem incompat and ro_compat bit fields to judge
whether the filesystem can be read or written.
For the read side only a scary warning is shown so far.
I'd love to abort mounting too, but I fear this will break some setups
where the driver works by chance.

Signed-off-by: Richard Weinberger <richard@nod.at>
23 months agolib: elf: Fix bootelf -p crash issue
JianfengA.Zhu@sony.com [Wed, 7 Aug 2024 02:20:07 +0000 (02:20 +0000)] 
lib: elf: Fix bootelf -p crash issue

Do not load and flush segments whose program header type is not
PT_LOAD to avoid crash happen.

bootelf will load all program headers and then flush cache.
Since PT_GNU_STACK start address is 0x0, crash happens when
it flushes PT_GNU_STACK area for target that memory cannot be
accessed from address 0x0.

crash call stack:
====================================================================
bootelf -p
|-> do_bootelf (cmd/elf.c)
 |-> load_elf_image_phdr (lib/elf.c) /* Load each program header */
  |-> for (i = 0; i < ehdr->e_phnum; ++i) { /* e_phnum is 2 */
   |-> flush_cache(start=0, size=0) (arch/arm/lib/cache.c) (*1)
    |-> flush_dcache_range(start=0, 0); (arch/arm/cpu/armv8/cache_v8.c)
     |-> __asm_flush_dcache_range(start=0, 0) (arch/arm/cpu/armv8/cache.S)
      |-> /* clean & invalidate data or unified cache */
      |-> dc      civac, x0       <<== crash happen x0:start=0
====================================================================
*1 Here the p_paddr and p_filesz of the GNU_STACK program are 0.

hello_world elf image program headers
====================================================================
Program Headers:
  Type           Offset             VirtAddr           PhysAddr
                 FileSiz            MemSiz              Flags  Align
  LOAD           0x0000000000000000 0x0000000080ff0000 0x0000000080ff0000
                 0x0000000000010390 0x0000000000010390  R E    0x10000
  GNU_STACK      0x0000000000000000 0x0000000000000000 0x0000000000000000
                 0x0000000000000000 0x0000000000000000  RW     0x10

 Section to Segment mapping:
  Segment Sections...
   00     .text .rodata .eh_frame
   01
====================================================================

bootelf -p crash error log
====================================================================
bootelf -p 0x80f00000
"Synchronous Abort" handler, esr 0x96000147, far 0x0
elr: 000000008f6035bc lr : 000000008f66ff54 (reloc)
elr: 00000000bd9475bc lr : 00000000bd9b3f54
x0 : 0000000000000000 x1 : 0000000000000000
x2 : 0000000000000040 x3 : 000000000000003f
x4 : 0000000081000390 x5 : 00000000bd12da98
x6 : 0000000000000010 x7 : 0000000000000010
x8 : 000000000000000a x9 : 00000000bd156b30
x10: 00000000bd1526e0 x11: 0000000000000000
x12: 0000000000000004 x13: 00000000bda00000
x14: 00000000ffffffff x15: 00000000bd12dab7
x16: 00000000bd957b14 x17: 0000000000000000
x18: 00000000bd13fd90 x19: 0000000080f000b0
x20: 0000000080f00000 x21: 0000000000000002
x22: 0000000000000000 x23: 0000000000000003
x24: 00000000bd9eacc0 x25: 0000000000000000
x26: 0000000000000000 x27: 0000000000000000
x28: 00000000bd154c80 x29: 00000000bd12da70

Code: d2800082 9ac32042 d1000443 8a230000 (d50b7e20)
====================================================================

Signed-off-by: Jianfeng Zhu <JianfengA.Zhu@sony.com>
Reviewed-by: Jacky Cao <Jacky.Cao@sony.com>
Reviewed-by: Toyama, Yoshihiro <Yoshihiro.Toyama@sony.com>
23 months agoarm: dts: keystone: Move to OF_UPSTREAM
Andrew Davis [Wed, 7 Aug 2024 14:56:51 +0000 (09:56 -0500)] 
arm: dts: keystone: Move to OF_UPSTREAM

Enable OF_UPSTREAM for Keystone boards. Remove DT files that
are now available in dts/upstream.

Signed-off-by: Andrew Davis <afd@ti.com>
Acked-by: Neha Malcom Francis <n-francis@ti.com>
23 months agopinmux: generic: Use ENOENT instead of ENOSYS
Sean Anderson [Sat, 10 Aug 2024 13:59:37 +0000 (09:59 -0400)] 
pinmux: generic: Use ENOENT instead of ENOSYS

ENOSYS should only be used when a subsystem is completely absent.
Convert its use in pinctrl-generic to ENOENT, which better reflects the
error condition (that a function/group/pin is missing).

Signed-off-by: Sean Anderson <seanga2@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
23 months agodefconfig: Add a config for AM335x High Security EVM with SPI Boot support
Andrew Davis [Wed, 7 Aug 2024 15:26:59 +0000 (10:26 -0500)] 
defconfig: Add a config for AM335x High Security EVM with SPI Boot support

Add a new defconfig file for the AM335x High Security EVM. This config
is specific for the case of SPI booting.

Signed-off-by: Andrew Davis <afd@ti.com>
23 months agoscripts/decodecode: update from Linux v6.10
Heinrich Schuchardt [Fri, 9 Aug 2024 18:22:33 +0000 (20:22 +0200)] 
scripts/decodecode: update from Linux v6.10

For decoding RISC-V dumps we need to update the script.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
23 months agoenv: remove vars that are not in default env
Ravi Minnikanti [Sun, 11 Aug 2024 18:44:15 +0000 (11:44 -0700)] 
env: remove vars that are not in default env

current env_set_default_vars() doesn't delete
var that are not in the imported env. hashtable
removes vars that are not in the imported
env but present in the current env only if H_NOCLEAR
flag is not set.

This change is to avoid passing H_NOCLEAR flag if
specific vars are passed to env_set_default_vars()

Without this change:
Marvell>> env default boot_mode
Marvell>>

With the change:
Marvell>> env default boot_mode
WARNING: 'boot_mode' not in imported env, deleting it!

Signed-off-by: Ravi Minnikanti <rminnikanti@marvell.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
23 months agosandbox: fix wget test failure after fixing wget issue
Yasuharu Shibata [Wed, 14 Aug 2024 12:41:07 +0000 (21:41 +0900)] 
sandbox: fix wget test failure after fixing wget issue

After applying the following patch, wget test on sandbox failed[1].

  Commit: cab7867cff ("net: wget: Support retransmission a dropped packet")

Here are two reasons why the test is failed and how to fix it:

1. tcp_ack is calculated by the wrong value.
   tcp_ack needs to be calculated by the received TCP payload size.
2. wget command may have a problem that HTTP response from server
   must be divided into more than two packets.
   In this commit, HTTP response is divided into two packets.

In addition, I fixed the HTTP response returned at the correct timing.

[1] https://lore.kernel.org/u-boot/CAFLszThEbk2Jr8OZ6Hj21wPSnJjgJhaDe037RqwHvwt1KjB3_A@mail.gmail.com/

Signed-off-by: Yasuharu Shibata <yasuharu.shibata@gmail.com>
Reported-by: Simon Glass <sjg@chromium.org>
23 months agosandbox: Enable wget command
Yasuharu Shibata [Wed, 14 Aug 2024 12:41:06 +0000 (21:41 +0900)] 
sandbox: Enable wget command

Enable this so that the tests run.

Fix a few warnings in the code so that CI passes.

Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Yasuharu Shibata <yasuharu.shibata@gmail.com>
23 months agocmd: led: fix led command usage for the LED_SW_BLINK enabled case
Mikhail Kshevetskiy [Wed, 14 Aug 2024 10:08:30 +0000 (14:08 +0400)] 
cmd: led: fix led command usage for the LED_SW_BLINK enabled case

CONFIG_LED_BLINK and CONFIG_LED_SW_BLINK can be defined independently.
Led blinking works if any (or both) of them is enabled. Unfortunately
the led command help does not display blinking option if only
CONFIG_LED_SW_BLINK is enabled. This is definitely wrong.

This patch fix an issue.

Signed-off-by: Mikhail Kshevetskiy <mikhail.kshevetskiy@iopsys.eu>
Reviewed-by: Simon Glass <sjg@chromium.org>
Acked-by: Alexander Dahl <ada@thorsis.com>
23 months agospi: soft_spi: Parse cs-gpios only if num-chipselects is not <0>
Michael Polyntsov [Wed, 31 Jul 2024 04:11:29 +0000 (08:11 +0400)] 
spi: soft_spi: Parse cs-gpios only if num-chipselects is not <0>

Some boards don't have chipselect lines for leds so cs-gpios is not
specified in the dts leading to probing error. Fix it by making
behavior similar to the one in Linux, parse num-chipselects and
if it is zero, ignore cs-gpios.

Signed-off-by: Michael Polyntsov <michael.polyntsov@iopsys.eu>
Signed-off-by: Mikhail Kshevetskiy <mikhail.kshevetskiy@iopsys.eu>
23 months agospi: soft_spi: fix miso gpio property name
Mikhail Kshevetskiy [Wed, 31 Jul 2024 04:11:28 +0000 (08:11 +0400)] 
spi: soft_spi: fix miso gpio property name

The patch fix a missprint introduced in commit 2e9fe73a883a ("spi: soft_spi:
Support the recommended soft spi properties").

Signed-off-by: Mikhail Kshevetskiy <mikhail.kshevetskiy@iopsys.eu>
Reviewed-by: Fabio Estevam <festevam@gmail.com>
23 months agomaintainers: Update list of maintainers for Corstone-1000
Hugues Kamba Mpiana [Tue, 13 Aug 2024 15:53:05 +0000 (16:53 +0100)] 
maintainers: Update list of maintainers for Corstone-1000

- Add new maintainer: Hugues KAMBA MPIANA
- Remove maintainer: Xueliang ZHONG
- Update contact information for current maintainer.

Signed-off-by: Hugues KAMBA MPIANA <hugues.kambampiana@arm.com>
23 months agoMerge tag 'u-boot-imx-master-20240813' of https://gitlab.denx.de/u-boot/custodians...
Tom Rini [Tue, 13 Aug 2024 16:10:29 +0000 (10:10 -0600)] 
Merge tag 'u-boot-imx-master-20240813' of https://gitlab.denx.de/u-boot/custodians/u-boot-imx

CI: https://source.denx.de/u-boot/custodians/u-boot-imx/-/pipelines/22014

- Convert tqma6q_mba6 to watchdog DM to fix reset.
- Convert tqma6q_mba6 to PMIC and I2C DM.
- Convert tqma6q_mba6 to OF_UPSTREAM.
- Do not print the board name twice on tqma6.
- Enable CMD_ERASEENV for imx8mm/mp Phytec boards.
- Add imx8ulp binman support.
- Fix imx8 build when CONFIG_IMX_BOOTAUX is set.

23 months agoMerge tag 'i2c-dm-updates-for-v2024-10-next' of https://source.denx.de/u-boot/custodi...
Tom Rini [Tue, 13 Aug 2024 16:09:45 +0000 (10:09 -0600)] 
Merge tag 'i2c-dm-updates-for-v2024-10-next' of https://source.denx.de/u-boot/custodians/u-boot-i2c into next

i2c DM updates for v2024.10 next

- Add series from Simon, which removes old content in i2c.h
  so that we can move towards having just the dm API.

  It does not remove any boards, but instead removes functionality.

23 months agotqma6: Do not print the board name twice
Fabio Estevam [Fri, 9 Aug 2024 17:58:09 +0000 (14:58 -0300)] 
tqma6: Do not print the board name twice

Currently, the devicetree model as well as the board variant name
are shown:
...
Model: TQ TQMa6S/DL on MBa6x
Board: TQMa6DL on a MBa6x
...

Unselect the CONFIG_DISPLAY_BOARDINFO option so that the
board name is printed only once in board_late_init() instead.

Signed-off-by: Fabio Estevam <festevam@denx.de>
23 months agotqma6: Convert to PMIC and I2C driver model
Fabio Estevam [Fri, 9 Aug 2024 17:58:08 +0000 (14:58 -0300)] 
tqma6: Convert to PMIC and I2C driver model

Currently, the power_init_board() function is not executed because
CONFIG_POWER_LEGACY is not selected.

Convert to PMIC driver model, which allows removing board I2C code in
favor of the I2C driver model.

Signed-off-by: Fabio Estevam <festevam@denx.de>
23 months agoimx6-tqma6: Convert to OF_UPSTREAM
Fabio Estevam [Fri, 9 Aug 2024 15:25:48 +0000 (12:25 -0300)] 
imx6-tqma6: Convert to OF_UPSTREAM

Instead of using the local imx6-tqma6 devicetree copies from U-Boot,
convert the imx6-tqma6 target to OF_UPSTREAM so that the upstream
kernel devicetrees can be used instead.

Signed-off-by: Fabio Estevam <festevam@denx.de>
23 months agotqma6q_mba6: Convert to watchdog driver model
Fabio Estevam [Fri, 9 Aug 2024 15:12:47 +0000 (12:12 -0300)] 
tqma6q_mba6: Convert to watchdog driver model

Commit 68dcbdd594d4 ("ARM: imx: Add weak default reset_cpu()") caused
the 'reset' command in U-Boot to not cause a board reset.

Fix it by switching to the watchdog driver model via sysreset, which
is the preferred method for implementing the watchdog reset.

Signed-off-by: Fabio Estevam <festevam@denx.de>
23 months agoimx: imx8: fix build when CONFIG_IMX_BOOTAUX is set
Max Krummenacher [Wed, 7 Aug 2024 13:39:11 +0000 (15:39 +0200)] 
imx: imx8: fix build when CONFIG_IMX_BOOTAUX is set

Use correct function name.

Fixes: e8cd1f60d964 ("imx: imx8: bootaux: Add i.MX8 M4 boot support")
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
23 months agoimx8ulp_evk: enable binman support
Gary Bisson [Mon, 5 Aug 2024 21:25:11 +0000 (23:25 +0200)] 
imx8ulp_evk: enable binman support

Enable binman support and add documentation for the
imx8ul-evk board.

Signed-off-by: Gary Bisson <bisson.gary@gmail.com>