[U-Boot] [PATCH] ARM: socfpga: Add boot0 hook to prevent SPL corruption

[Chin Liang See]

On Rab, 2016-11-16 at 23:39 +0100, Marek Vasut wrote:
> On 11/16/2016 08:27 PM, Dinh Nguyen wrote:
> > 
> > 
> > 
> > On 11/16/2016 10:54 AM, Marek Vasut wrote:
> > > 
> > > Valid Altera SoCFPGA preloader image must contain special data at
> > > offsets 0x40, 0x44, 0x48 and valid instructions at address 0x4c
> > > or
> > > 0x50. These addresses are by default used by U-Boot's vector
> > > table
> > > and a piece of reset handler, thus a valid preloader corrupts
> > > those
> > > addresses slightly. While this works most of the time, this can
> > > and
> > > does prevent the board from rebooting sometimes and triggering
> > > this
> > > issue may even depend on compiler.
> > > 
> > > The problem is that when SoCFPGA performs warm reset, it checks
> > > the
> > > addresses 0x40..0x4b in SRAM for a valid preloader signature and
> > > header checksum. If those are found, it jumps to address 0x4c or
> > > 0x50 (this is unclear). These addresses are populated by the
> > > first
> > > few instructions of arch/arm/cpu/armv7/start.S:
> > > 
> > > ffff0040 <data_abort>:
> > > ffff0040:       ebfffffe        bl      ffff0040 <data_abort>
> > > 
> > > ffff0044 <reset>:
> > > ffff0044:       ea000012        b       ffff0094
> > > <save_boot_params>
> > > 
> > > ffff0048 <save_boot_params_ret>:
> > > ffff0048:       e10f0000        mrs     r0, CPSR
> > > ffff004c:       e200101f        and     r1, r0, #31
> > > ffff0050:       e331001a        teq     r1, #26
> > > 
> > > Without this patch, the CPU will enter the code at 0xffff004c or
> > > 0xffff0050 , at which point the value of r0 and r1 registers is
> > > undefined. Moreover, jumping directly to the preloader entry
> > > point
> > > at address 0xffff0000 will also fail, because address 0xffff004.
> > > is invalid and contains the preloader magic.
> > > 
> > > Add BOOT0 hook which reserves the area at offset 0x40..0x5f and
> > > populates offset 0x50 with jump to the entry point. This way, the
> > > preloader signature is stored in reserved space and can not
> > > corrupt
> > > the SPL code.
> > > 
> > > Signed-off-by: Marek Vasut <marex at denx.de>
> > > Cc: Chin Liang See <clsee at altera.com>
> > > Cc: Dinh Nguyen <dinguyen at opensource.altera.com>
> > > Cc: Stefan Roese <sr at denx.de>
> > > ---
> > >  arch/arm/Kconfig                           |  1 +
> > >  arch/arm/mach-socfpga/include/mach/boot0.h | 28
> > > ++++++++++++++++++++++++++++
> > >  2 files changed, 29 insertions(+)
> > >  create mode 100644 arch/arm/mach-socfpga/include/mach/boot0.h
> > With this patch, I can properly do a soft reset on an Atlas DE0
> > Nano board.
> > 
> > So feel free to add:
> > 
> > Tested-by: Dinh Nguyen <dinguyen at opensource.altera.com>
> That's great, thanks for checking ! You probably want to propagate
> this
> fix to your downstream U-Boot mutation and also that MPL loader.
> 

Great, this fix in time as we noticed this issue early of this week. We
were testing out the Arria10 SoC SPL. While for downstream, we already
cater that but not gracefully as we modified the start.s :)

THanks
Chin Liang

> --
> Best regards,
> Marek Vasut
> 
> ________________________________
> 
> Confidentiality Notice.
> This message may contain information that is confidential or
> otherwise protected from disclosure. If you are not the intended
> recipient, you are hereby notified that any use, disclosure,
> dissemination, distribution, or copying of this message, or any
> attachments, is strictly prohibited. If you have received this
> message in error, please advise the sender by reply e-mail, and
> delete the message and any attachments. Thank you.