Skip to contentSkip to conversation

p2371-2180: Build position independent binary

PATCH New Thierry Reding 2 participants 6 messages Mar 8, 2019 – Mar 20, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 p2371-2180: Build position independent binary

From: Thierry Reding <[email protected]>

In order to support chainloading of U-Boot by an earlier bootloader, make sure the binary is position independent, so that the earlier boot- loader can relocate it if necessary.

---
 configs/p2371-2180_defconfig | 1 +
 1 file changed, 1 insertion(+)
configs/p2371-2180_defconfig +1 -0
index b66459e379ac..8d7cf3fb5346 100644
@@ -1,6 +1,7 @@
 CONFIG_ARM=y
 CONFIG_TEGRA=y
 CONFIG_SYS_TEXT_BASE=0x80110000
+CONFIG_POSITION_INDEPENDENT=y
 CONFIG_TEGRA210=y
 CONFIG_TARGET_P2371_2180=y
 CONFIG_NR_DRAM_BANKS=2
Signature · 1 line
2.20.1
#1 lore raw

PATCH 1/1 p2371-2180: Build position independent binary

Quoted 5 lines · Thierry Reding

From: Thierry Reding <[email protected]>

In order to support chainloading of U-Boot by an earlier bootloader, make sure the binary is position independent, so that the earlier boot- loader can relocate it if necessary.

Why not enable this for all 64-bit Tegra? They're all booted the exact same way at least with recent L4T builds.

Also, U-Boot is typically linked to the address that cboot loads it to, and cboot typically always loads to precisely that address. I'm not sure why this patch is required.

That said, it don't think it harms anything, so I'm fine with it being applied.

#2 lore raw

PATCH 1/1 p2371-2180: Build position independent binary

Quoted 9 lines · Stephen Warren
Quoted 5 lines · Thierry Reding

From: Thierry Reding <[email protected]>

In order to support chainloading of U-Boot by an earlier bootloader, make sure the binary is position independent, so that the earlier boot- loader can relocate it if necessary.

Why not enable this for all 64-bit Tegra? They're all booted the exact same way at least with recent L4T builds.

Yeah, I think that would make sense.

Quoted 3 lines

Also, U-Boot is typically linked to the address that cboot loads it to, and cboot typically always loads to precisely that address. I'm not sure why this patch is required.

I encountered this issue when I was trying to chainload U-Boot from cboot on Jetson TX1. It seems like your above comment is no longer true, though I suppose that could just mean that the link address for U-Boot has become stale?

Quoted 2 lines

That said, it don't think it harms anything, so I'm fine with it being applied.

I suppose there's a bit of extra code to do the indirect jumps, but overall U-Boot seems to work well and not noticeably slower if this is enabled. Might be nice for extra flexibility and to avoid any surprises if we ever end up loading U-Boot to a location other than where it was liked to.

My understanding is that this could happen on Tegra186 if cboot detects bad memory blocks in the area where U-Boot was meant to be loaded to. I guess this doesn't apply to earlier chips, but perhaps it's good to have it there for consistency anyway.

Thierry
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190319/384b1e35/attachment.sig>

#3 lore raw

PATCH 1/1 p2371-2180: Build position independent binary

Quoted 13 lines · Thierry Reding
Quoted 9 lines · Stephen Warren
Quoted 5 lines · Thierry Reding

From: Thierry Reding <[email protected]>

In order to support chainloading of U-Boot by an earlier bootloader, make sure the binary is position independent, so that the earlier boot- loader can relocate it if necessary.

Why not enable this for all 64-bit Tegra? They're all booted the exact same way at least with recent L4T builds.

Yeah, I think that would make sense.

Quoted 8 lines
Quoted 3 lines

Also, U-Boot is typically linked to the address that cboot loads it to, and cboot typically always loads to precisely that address. I'm not sure why this patch is required.

I encountered this issue when I was trying to chainload U-Boot from cboot on Jetson TX1. It seems like your above comment is no longer true, though I suppose that could just mean that the link address for U-Boot has become stale?

Looks like the upstream CONFIG_SYS_TEXT_BASE is indeed stale relative to the latest L4T builds:

Upstream:
Jetson TX1: CONFIG_SYS_TEXT_BASE=0x80110000
Jetson TX2: CONFIG_SYS_TEXT_BASE=0x80080000

L4T r32.1:
Jetson TX1: #define CONFIG_SYS_TEXT_BASE 0x80080000
Jetson TX2: #define CONFIG_SYS_TEXT_BASE 0x80080000

Quoted 9 lines
Quoted 2 lines

That said, it don't think it harms anything, so I'm fine with it being applied.

I suppose there's a bit of extra code to do the indirect jumps, but overall U-Boot seems to work well and not noticeably slower if this is enabled. Might be nice for extra flexibility and to avoid any surprises if we ever end up loading U-Boot to a location other than where it was liked to.

Quoted 4 lines

My understanding is that this could happen on Tegra186 if cboot detects bad memory blocks in the area where U-Boot was meant to be loaded to. I guess this doesn't apply to earlier chips, but perhaps it's good to have it there for consistency anyway.

Yes, this could happen on Jetson TX2i at least.

#4 lore raw

PATCH 1/1 p2371-2180: Build position independent binary

Quoted 26 lines · Stephen Warren
Quoted 13 lines · Thierry Reding
Quoted 9 lines · Stephen Warren
Quoted 5 lines · Thierry Reding

From: Thierry Reding <[email protected]>

In order to support chainloading of U-Boot by an earlier bootloader, make sure the binary is position independent, so that the earlier boot- loader can relocate it if necessary.

Why not enable this for all 64-bit Tegra? They're all booted the exact same way at least with recent L4T builds.

Yeah, I think that would make sense.

Quoted 8 lines
Quoted 3 lines

Also, U-Boot is typically linked to the address that cboot loads it to, and cboot typically always loads to precisely that address. I'm not sure why this patch is required.

I encountered this issue when I was trying to chainload U-Boot from cboot on Jetson TX1. It seems like your above comment is no longer true, though I suppose that could just mean that the link address for U-Boot has become stale?

Looks like the upstream CONFIG_SYS_TEXT_BASE is indeed stale relative to the latest L4T builds:

Quoted 7 lines

Upstream:
Jetson TX1: CONFIG_SYS_TEXT_BASE=0x80110000
Jetson TX2: CONFIG_SYS_TEXT_BASE=0x80080000

L4T r32.1:
Jetson TX1: #define CONFIG_SYS_TEXT_BASE 0x80080000
Jetson TX2: #define CONFIG_SYS_TEXT_BASE 0x80080000

Okay, let me fix that up while at it. I think the 0x80110000 text base was something that we cargo-culted from 32-bit ARM boards. I vaguely recall that this might have been related to the SPL split in some way, but I can't find anything to corroborate that.

Quoted 9 lines
Quoted 9 lines
Quoted 2 lines

That said, it don't think it harms anything, so I'm fine with it being applied.

I suppose there's a bit of extra code to do the indirect jumps, but overall U-Boot seems to work well and not noticeably slower if this is enabled. Might be nice for extra flexibility and to avoid any surprises if we ever end up loading U-Boot to a location other than where it was liked to.

Quoted 6 lines
Quoted 4 lines

My understanding is that this could happen on Tegra186 if cboot detects bad memory blocks in the area where U-Boot was meant to be loaded to. I guess this doesn't apply to earlier chips, but perhaps it's good to have it there for consistency anyway.

Yes, this could happen on Jetson TX2i at least.

Let me send out a proposal than makes all 64-bit Tegra builds position independent and throws in the TEXT_BASE changes discussed above.

Thierry
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190320/0436c534/attachment-0001.sig>

#5 lore raw

PATCH 1/1 p2371-2180: Build position independent binary

Quoted 31 lines · Thierry Reding
Quoted 30 lines · Stephen Warren
Quoted 13 lines · Thierry Reding
Quoted 9 lines · Stephen Warren
Quoted 5 lines · Thierry Reding

From: Thierry Reding <[email protected]>

In order to support chainloading of U-Boot by an earlier bootloader, make sure the binary is position independent, so that the earlier boot- loader can relocate it if necessary.

Why not enable this for all 64-bit Tegra? They're all booted the exact same way at least with recent L4T builds.

Yeah, I think that would make sense.

Quoted 8 lines
Quoted 3 lines

Also, U-Boot is typically linked to the address that cboot loads it to, and cboot typically always loads to precisely that address. I'm not sure why this patch is required.

I encountered this issue when I was trying to chainload U-Boot from cboot on Jetson TX1. It seems like your above comment is no longer true, though I suppose that could just mean that the link address for U-Boot has become stale?

Looks like the upstream CONFIG_SYS_TEXT_BASE is indeed stale relative to the latest L4T builds:

Upstream:
Jetson TX1: CONFIG_SYS_TEXT_BASE=0x80110000
Jetson TX2: CONFIG_SYS_TEXT_BASE=0x80080000

Quoted 8 lines
Quoted 3 lines

L4T r32.1:
Jetson TX1: #define CONFIG_SYS_TEXT_BASE 0x80080000
Jetson TX2: #define CONFIG_SYS_TEXT_BASE 0x80080000

Okay, let me fix that up while at it. I think the 0x80110000 text base was something that we cargo-culted from 32-bit ARM boards. I vaguely recall that this might have been related to the SPL split in some way, but I can't find anything to corroborate that.

I believe it was accurate for the first L4T release to support Tegra210, which chain-loaded U-Boot from nvtboot-cpu rather than chain-loading it from cboot. But, I could be wrong.