probe_ram_size_by_alias() detects whether a probe address still aliases
a lower address by writing through one address and reading through the
other.
On i.MX95 this occasionally reported a false non-alias when the alias
read happened immediately after the write.
A memory barrier alone, mb(), was tested but did not make the failure go
away. This suggests that ordering the CPU accesses is not sufficient for
this probe, likely because the issue is in the path to the memory
controller rather than in the core itself.
Read the written address back before checking the alias address. This
appears to force the write to become observable at the probe address
before using the alias read to decide whether the tested address range
exists.
If the readback does not match the written pattern, restore the saved
value and continue with the next check. This keeps the probe robust for
addresses that do not reliably retain the test pattern.
Fixes: 0977448b45 ("common: memsize: add RAM size probe based on alias detection")
Signed-off-by: Emanuele Ghidoli <emanuele.ghidoli@toradex.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Add probe_ram_size_by_alias() to detect RAM size by checking whether a
write to one address aliases to another address.
Compared to get_ram_size(), this function allows the caller to:
- limit probing to a small set of required accesses
- avoid touching reserved or already used memory regions
- handle non-linear alias patterns
On the iMX95 SoC, when used with LPDDR5, accesses beyond the end of an 8GB DDR
configuration do not alias to the expected linear wrap-around addresses.
Instead, the aliased addresses appear to follow a pattern related to the
DDRC bank and bank-group addresses mapping. Experimentally, the observed
pattern is:
Write Read
y00000000 -> x0001c000
y00004000 -> x00018000
y00008000 -> x00014000
y0000c000 -> x00010000
y00010000 -> x0000c000
y00014000 -> x00008000
y00018000 -> x00004000
y0001c000 -> x00000000
This helper makes it possible to probe RAM size by explicitly specifying
the probed address and the expected alias address for each size check.
Signed-off-by: Emanuele Ghidoli <emanuele.ghidoli@toradex.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
The get_ram_size() function fails to restore the original RAM data when
the data cache is enabled. This issue was observed on an AM625 R5 SPL
with 512MB of RAM and is a regression that became visible with
commit bc07851897 ("board: ti: Pull redundant DDR functions to a common
location and Fixup DDR size when ECC is enabled").
Observed boot failure messages:
Warning: Did not detect image signing certificate. Skipping authentication to prevent boot failure. This will fail on Security Enforcing(HS-SE) devices
Authentication passed
Starting ATF on ARM64 core...
The system then hangs. This indicates that without a data cache flush,
data in the cache is not coherent with RAM, preventing the system from
booting. This was verified by printing the content of this address when
the issue occurs.
Add a data cache flush after each restore operation to resolve this
issue.
Fixes: bc07851897 ("board: ti: Pull redundant DDR functions to a common location and Fixup DDR size when ECC is enabled")
Fixes: 1c64b98c1e ("common/memsize.c: Fix get_ram_size() when cache is enabled")
Signed-off-by: Stefan Eichenberger <stefan.eichenberger@toradex.com>
Reviewed-by: Emanuele Ghidoli <emanuele.ghidoli@toradex.com>
Tested-by: Francesco Dolcini <francesco.dolcini@toradex.com> # Toradex Verdin AM62
When the data cache (dcache) is disabled, calling related
status functions can lead to compilation errors due to
undefined references.
Adding a !CONFIG_IS_ENABLED(SYS_DCACHE_OFF) check before
invoking dcache_status() (used in common/memsize.c:get_ram_size())
and mmu_status() (from arch/arm/include/asm/io.h).
Without this check, builds with dcache disabled will fail to compile.
Signed-off-by: Boon Khai Ng <boon.khai.ng@altera.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
As part of bringing the master branch back in to next, we need to allow
for all of these changes to exist here.
Reported-by: Jonas Karlman <jonas@kwiboo.se>
Signed-off-by: Tom Rini <trini@konsulko.com>
When bringing in the series 'arm: dts: am62-beagleplay: Fix Beagleplay
Ethernet"' I failed to notice that b4 noticed it was based on next and
so took that as the base commit and merged that part of next to master.
This reverts commit c8ffd1356d, reversing
changes made to 2ee6f3a5f7.
Reported-by: Jonas Karlman <jonas@kwiboo.se>
Signed-off-by: Tom Rini <trini@konsulko.com>
Ensure that every write is flushed to memory and afterward reads are
from memory.
Since the algorithm rely on the fact that accessing to not existent
memory lead to write at addr / 2 without this modification accesses
to aliased (not physically present) addresses are cached and
wrong size is returned.
This was discovered while working on a TI AM625 based board
where cache is normally enabled, see commit c02712a748 ("arm: mach-k3: Enable dcache in SPL").
Signed-off-by: Emanuele Ghidoli <emanuele.ghidoli@toradex.com>
Signed-off-by: Francesco Dolcini <francesco.dolcini@toradex.com>
This reverts commit 777aaaa706 ("common/memsize.c: Fix
get_effective_memsize() to check for overflow") for non-mpc85xx platforms.
The changes to this generic function, which is intended to help with
32bit platforms with large amounts of memory has unintended side effects
(which in turn lead to boot failures) on other platforms which were
previously functional.
For now do overflow check only for powerpc mpc85xx platform. It is needed
to prevent crashing of P1/P2 boards with 4GB DDR module in 32-bit mode.
Fixes: 777aaaa706 ("common/memsize.c: Fix get_effective_memsize() to check for overflow")
Signed-off-by: Pali Rohár <pali@kernel.org>
Ensure that top of RAM can be represented by phys_size_t type. If RAM is
too large or RAM base address is too upper then limit RAM size to prevent
address space overflow.
Signed-off-by: Pali Rohár <pali@kernel.org>
CONFIG_MAX_MEM_MAPPED when defined specifies upper memory mapped limit.
So check for it always, and not only when CONFIG_VERY_BIG_RAM is defined.
Signed-off-by: Pali Rohár <pali@kernel.org>
Move this out of the common header and include it only where needed. In
a number of cases this requires adding "struct udevice;" to avoid adding
another large header or in other cases replacing / adding missing header
files that had been pulled in, very indirectly. Finally, we have a few
cases where we did not need to include <asm/global_data.h> at all, so
remove that include.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Rini <trini@konsulko.com>
In ARM 64-bits, memory size can be supported is more than 4GB,
hence increasing save array is needed to cope with testing larger memory.
Signed-off-by: Tien Fong Chee <tien.fong.chee@intel.com>
When U-Boot started using SPDX tags we were among the early adopters and
there weren't a lot of other examples to borrow from. So we picked the
area of the file that usually had a full license text and replaced it
with an appropriate SPDX-License-Identifier: entry. Since then, the
Linux Kernel has adopted SPDX tags and they place it as the very first
line in a file (except where shebangs are used, then it's second line)
and with slightly different comment styles than us.
In part due to community overlap, in part due to better tag visibility
and in part for other minor reasons, switch over to that style.
This commit changes all instances where we have a single declared
license in the tag as both the before and after are identical in tag
contents. There's also a few places where I found we did not have a tag
and have introduced one.
Signed-off-by: Tom Rini <trini@konsulko.com>
For the last return of function get_ram_size(),
when size=maxsize, restore the base address (*base)
content.
Add comment for the remaining case to avoid regression:
this case is already correctly handled.
Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>
Save the content of the base address in a new variable
in stack (save_base) to prepare restore this content.
This patch don't modified code behavior and stack usage.
Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>
This commit breaks bootup on sunxi boards, the get stuck
when running the main u-boot binary at:
CPU: Allwinner H3 (SUN8I)
I2C: ready
DRAM:
This reverts commit 8e7cba048b.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
The case of memory of size 0 is not that different from a memory of any other
size, so we remove the duplicate code and treat the small differences when it
is the case.
Signed-off-by: Eddy Petrișor <eddy.petrisor@gmail.com>
This function has been around for powerpc. It is used for systems with
memory more than CONFIG_MAX_MEM_MAPPED. In case of non-contiguous memory,
this feature can limit U-boot to one block without going over the limit.
Signed-off-by: York Sun <yorksun@freescale.com>
Acked-by: Albert ARIBAUD <albert.u.boot@aribaud.net>
After further investigation this change is not correct and results in
get_ram_size returning incorrectly large values. Reported first by
Gerlando Falauto.
This reverts commit b8496cced8.
Signed-off-by: Tom Rini <trini@ti.com>
The original memory sizing code in get_ram_size clobbers the word
at the base address, but forgets to restore it.
Signed-off-by: Iwo Mergler <Iwo.Mergler@netcommwireless.com>
Checkpatch.pl complains about the volatile qualifier in calls to
get_ram_size(). Remove this qualifier in the prototype and in the
calls where it is useless, and leave it only in the function body
where it is needed.
Signed-off-by: Albert ARIBAUD <albert.u.boot@aribaud.net>
