Skip to contentSkip to conversation

spl: full-featured heap cleanups

PATCH New Simon Goldschmidt 4 participants 18 messages Apr 1, 2019 – Apr 29, 2019

lore
Share & tools
lore
#0 lore raw

v4 cover spl: full-featured heap cleanups

Since clearing BSS before board_init_r is clearly not accepted, let's just fix 2 bugs in dlmalloc and I'll handle the socfpga_a10 cleanup in a different way (see discussions in v1 to v3 of this series...).

Changes in v4:
- dumped clearing BSS before SPL board_init_f, only real bugfixes remain

Changes in v3:
- fixed summary ("stack" -> "heap")
- enable CONFIG_SPL_CLEAR_BSS_F for socfpga_arria10 using full malloc early in SPL
- rebased

Simon Goldschmidt (2):
dlmalloc: fix malloc range at end of ram
dlmalloc: be compatible to tiny printf

 common/dlmalloc.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
Signature · 1 line
2.17.1
#1 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

---

Changes in v4: None
Changes in v3: None

 common/dlmalloc.c | 4 ++++
 1 file changed, 4 insertions(+)
common/dlmalloc.c +4 -0
index edaad299bb..51d3bd671a 100644
@@ -603,6 +603,10 @@ void mem_malloc_init(ulong start, ulong size)
 	mem_malloc_start = start;
 	mem_malloc_end = start + size;
 	mem_malloc_brk = start;
+	if (start && size && !mem_malloc_end) {
+		/* overflow: malloc area is at end of address range */
+		mem_malloc_end--;
+	}
 
 	debug("using memory %#lx-%#lx for malloc()\n", mem_malloc_start,
 	      mem_malloc_end);
Signature · 1 line
2.17.1
#2 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Quoted 7 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

Quoted 8 lines

---

Changes in v4: None
Changes in v3: None

 common/dlmalloc.c | 4 ++++
 1 file changed, 4 insertions(+)
#3 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Quoted 12 lines · Simon Glass
Quoted 10 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

---

Quoted 7 lines
Quoted 5 lines

Changes in v4: None
Changes in v3: None

 common/dlmalloc.c | 4 ++++
 1 file changed, 4 insertions(+)

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190424/4c2bd12e/attachment.sig>
#4 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Quoted 17 lines · Tom Rini
Quoted 15 lines · Simon Glass
Quoted 13 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

---

Changes in v4: None
Changes in v3: None

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines
 common/dlmalloc.c | 4 ++++
 1 file changed, 4 insertions(+)

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Ehrm, ok, so how do we proceed?

Regards,
Simon

#5 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Quoted 21 lines · Simon Goldschmidt
Quoted 20 lines · Tom Rini
Quoted 18 lines · Simon Glass
Quoted 15 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

---

Changes in v4: None
Changes in v3: None

 common/dlmalloc.c | 4 ++++
 1 file changed, 4 insertions(+)
Quoted 6 lines
Quoted 4 lines
Quoted 1 line

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Ehrm, ok, so how do we proceed?

A good question. Take a look at spl/u-boot-spl.map on smartweb and see if, of the malloc functions it doesn't discard there's something that maybe could be optimized somewhere?

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190424/998a9a69/attachment.sig>
#6 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Hi,

Quoted 25 lines · Tom Rini
Quoted 23 lines · Simon Goldschmidt
Quoted 22 lines · Tom Rini
Quoted 19 lines · Simon Glass
Quoted 7 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

Quoted 8 lines

---

Changes in v4: None
Changes in v3: None

 common/dlmalloc.c | 4 ++++
 1 file changed, 4 insertions(+)
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Ehrm, ok, so how do we proceed?

A good question. Take a look at spl/u-boot-spl.map on smartweb and see if, of the malloc functions it doesn't discard there's something that maybe could be optimized somewhere?

I wonder if we should have a Kconfig option like SPL_CHECKS which enables these sorts of minor checks, which may only fix one board at the cost of code size?

Then it could be enabled by default, but disabled on this board?

Regards,
Simon

#7 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Quoted 38 lines · Simon Glass

Hi,

Quoted 25 lines · Tom Rini
Quoted 23 lines · Simon Goldschmidt
Quoted 22 lines · Tom Rini
Quoted 19 lines · Simon Glass
Quoted 7 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

Quoted 8 lines

---

Changes in v4: None
Changes in v3: None

 common/dlmalloc.c | 4 ++++
 1 file changed, 4 insertions(+)
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Ehrm, ok, so how do we proceed?

A good question. Take a look at spl/u-boot-spl.map on smartweb and see if, of the malloc functions it doesn't discard there's something that maybe could be optimized somewhere?

Quoted 5 lines

I wonder if we should have a Kconfig option like SPL_CHECKS which enables these sorts of minor checks, which may only fix one board at the cost of code size?

Then it could be enabled by default, but disabled on this board?

For a bigger change, this might be an idea, but for a change that I can cut down to 16 or even 8 bytes code size increasement, I don't think having a new option would be good.

Anyway, I just tried at work and I don't get the overflow. Tom, which gcc are you using to get the size error? It works for me on Debian 9 but doesn't work with Ubuntu (both times, default cross compiler toolchain installed).

Regards,
Simon

#8 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Quoted 45 lines · Simon Goldschmidt
Quoted 38 lines · Simon Glass

Hi,

Quoted 25 lines · Tom Rini
Quoted 23 lines · Simon Goldschmidt
Quoted 22 lines · Tom Rini
Quoted 19 lines · Simon Glass
Quoted 7 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

Quoted 8 lines

---

Changes in v4: None
Changes in v3: None

 common/dlmalloc.c | 4 ++++
 1 file changed, 4 insertions(+)
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Ehrm, ok, so how do we proceed?

A good question. Take a look at spl/u-boot-spl.map on smartweb and see if, of the malloc functions it doesn't discard there's something that maybe could be optimized somewhere?

Quoted 5 lines

I wonder if we should have a Kconfig option like SPL_CHECKS which enables these sorts of minor checks, which may only fix one board at the cost of code size?

Then it could be enabled by default, but disabled on this board?

Quoted 7 lines

For a bigger change, this might be an idea, but for a change that I can cut down to 16 or even 8 bytes code size increasement, I don't think having a new option would be good.

Anyway, I just tried at work and I don't get the overflow. Tom, which gcc are you using to get the size error? It works for me on Debian 9 but doesn't work with Ubuntu (both times, default cross compiler toolchain installed).

I'm using the gcc-7.3 from kernel.org that we use in travis/etc.

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190425/8a2f8ce9/attachment.sig>
#9 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Am 25.04.2019 um 12:50 schrieb Tom Rini:

Quoted 50 lines · Simon Goldschmidt
Quoted 49 lines · Simon Glass
Quoted 38 lines

Hi,

Quoted 25 lines · Tom Rini
Quoted 23 lines · Simon Goldschmidt
Quoted 22 lines · Tom Rini
Quoted 19 lines · Simon Glass
Quoted 7 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

Quoted 8 lines

---

Changes in v4: None
Changes in v3: None

  common/dlmalloc.c | 4 ++++
  1 file changed, 4 insertions(+)
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Ehrm, ok, so how do we proceed?

A good question. Take a look at spl/u-boot-spl.map on smartweb and see if, of the malloc functions it doesn't discard there's something that maybe could be optimized somewhere?

Quoted 5 lines

I wonder if we should have a Kconfig option like SPL_CHECKS which enables these sorts of minor checks, which may only fix one board at the cost of code size?

Then it could be enabled by default, but disabled on this board?

For a bigger change, this might be an idea, but for a change that I can cut down to 16 or even 8 bytes code size increasement, I don't think having a new option would be good.

Quoted 5 lines
Quoted 3 lines

Anyway, I just tried at work and I don't get the overflow. Tom, which gcc are you using to get the size error? It works for me on Debian 9 but doesn't work with Ubuntu (both times, default cross compiler toolchain installed).

I'm using the gcc-7.3 from kernel.org that we use in travis/etc.

Ok, so I have gcc-7.3 on my Ubuntu machine as well. I don't know why 6.3 seems to produce smaller binaries (I thought they were getting smaller with new versions, not larger).

However, I've stripped down that patch to +8 Bytes only and sent v5.

Regards,
Simon

#10 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Hello Simon,

Am 25.04.2019 um 21:24 schrieb Simon Goldschmidt:

Quoted 55 lines

Am 25.04.2019 um 12:50 schrieb Tom Rini:

Quoted 54 lines · Simon Goldschmidt
Quoted 52 lines · Simon Glass
Quoted 38 lines

Hi,

Quoted 25 lines · Tom Rini
Quoted 23 lines · Simon Goldschmidt
Quoted 22 lines · Tom Rini
Quoted 19 lines · Simon Glass
Quoted 7 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

Quoted 8 lines

---

Changes in v4: None
Changes in v3: None

  common/dlmalloc.c | 4 ++++

  1 file changed, 4 insertions(+)
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Ehrm, ok, so how do we proceed?

A good question.  Take a look at spl/u-boot-spl.map on smartweb and see if, of the malloc functions it doesn't discard there's something that maybe could be optimized somewhere?

Quoted 5 lines

I wonder if we should have a Kconfig option like SPL_CHECKS which enables these sorts of minor checks, which may only fix one board at the cost of code size?

Then it could be enabled by default, but disabled on this board?

For a bigger change, this might be an idea, but for a change that I can cut down to 16 or even 8 bytes code size increasement, I don't think having a new option would be good.

Anyway, I just tried at work and I don't get the overflow. Tom, which gcc are you using to get the size error? It works for me on Debian 9 but doesn't work with Ubuntu (both times, default cross compiler toolchain installed).

Quoted 6 lines
Quoted 1 line

I'm using the gcc-7.3 from kernel.org that we use in travis/etc.

Ok, so I have gcc-7.3 on my Ubuntu machine as well. I don't know why 6.3 seems to produce smaller binaries (I thought they were getting smaller with new versions, not larger).

However, I've stripped down that patch to +8 Bytes only and sent v5.

Thanks!

Sorry for digging so late in, but I was on vacation...

Hmm.. the smartweb board has only 4k sram for SPL, and I have no chance to convert it to DM to get rid of some compiler warnings ...

I am unsure what to do now with this hardware ...

bye,
Heiko

Signature · 3 lines
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-52   Fax: +49-8142-66989-80   Email: hs at denx.de
#11 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Hello Heiko,

Quoted 66 lines · Heiko Schocher

Hello Simon,

Am 25.04.2019 um 21:24 schrieb Simon Goldschmidt:

Quoted 55 lines

Am 25.04.2019 um 12:50 schrieb Tom Rini:

Quoted 54 lines · Simon Goldschmidt
Quoted 52 lines · Simon Glass
Quoted 38 lines

Hi,

Quoted 25 lines · Tom Rini
Quoted 23 lines · Simon Goldschmidt
Quoted 22 lines · Tom Rini
Quoted 19 lines · Simon Glass
Quoted 7 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

Quoted 8 lines

---

Changes in v4: None
Changes in v3: None

  common/dlmalloc.c | 4 ++++
  1 file changed, 4 insertions(+)
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Ehrm, ok, so how do we proceed?

A good question. Take a look at spl/u-boot-spl.map on smartweb and see if, of the malloc functions it doesn't discard there's something that maybe could be optimized somewhere?

Quoted 5 lines

I wonder if we should have a Kconfig option like SPL_CHECKS which enables these sorts of minor checks, which may only fix one board at the cost of code size?

Then it could be enabled by default, but disabled on this board?

For a bigger change, this might be an idea, but for a change that I can cut down to 16 or even 8 bytes code size increasement, I don't think having a new option would be good.

Anyway, I just tried at work and I don't get the overflow. Tom, which gcc are you using to get the size error? It works for me on Debian 9 but doesn't work with Ubuntu (both times, default cross compiler toolchain installed).

Quoted 6 lines
Quoted 1 line

I'm using the gcc-7.3 from kernel.org that we use in travis/etc.

Ok, so I have gcc-7.3 on my Ubuntu machine as well. I don't know why 6.3 seems to produce smaller binaries (I thought they were getting smaller with new versions, not larger).

However, I've stripped down that patch to +8 Bytes only and sent v5.

Quoted 8 lines

Thanks!

Sorry for digging so late in, but I was on vacation...

Hmm.. the smartweb board has only 4k sram for SPL, and I have no chance to convert it to DM to get rid of some compiler warnings ...

I am unsure what to do now with this hardware ...

And things even get worse: as I wrote in the other thread, after updating to Ubuntu 19.04 as build system, I get gcc 8.3 as cross compiler and smartweb fails to build with that compiler (as the SPL binary is exactly 4k now).

Regards,
Simon

#12 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Quoted 73 lines · Simon Goldschmidt

Hello Heiko,

Quoted 70 lines · Heiko Schocher

Hello Simon,

Am 25.04.2019 um 21:24 schrieb Simon Goldschmidt:

Quoted 55 lines

Am 25.04.2019 um 12:50 schrieb Tom Rini:

Quoted 54 lines · Simon Goldschmidt
Quoted 52 lines · Simon Glass
Quoted 38 lines

Hi,

Quoted 25 lines · Tom Rini
Quoted 23 lines · Simon Goldschmidt
Quoted 22 lines · Tom Rini
Quoted 19 lines · Simon Glass
Quoted 7 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

Quoted 8 lines

---

Changes in v4: None
Changes in v3: None

  common/dlmalloc.c | 4 ++++
  1 file changed, 4 insertions(+)
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Ehrm, ok, so how do we proceed?

A good question. Take a look at spl/u-boot-spl.map on smartweb and see if, of the malloc functions it doesn't discard there's something that maybe could be optimized somewhere?

Quoted 5 lines

I wonder if we should have a Kconfig option like SPL_CHECKS which enables these sorts of minor checks, which may only fix one board at the cost of code size?

Then it could be enabled by default, but disabled on this board?

For a bigger change, this might be an idea, but for a change that I can cut down to 16 or even 8 bytes code size increasement, I don't think having a new option would be good.

Anyway, I just tried at work and I don't get the overflow. Tom, which gcc are you using to get the size error? It works for me on Debian 9 but doesn't work with Ubuntu (both times, default cross compiler toolchain installed).

Quoted 6 lines
Quoted 1 line

I'm using the gcc-7.3 from kernel.org that we use in travis/etc.

Ok, so I have gcc-7.3 on my Ubuntu machine as well. I don't know why 6.3 seems to produce smaller binaries (I thought they were getting smaller with new versions, not larger).

However, I've stripped down that patch to +8 Bytes only and sent v5.

Thanks!

Sorry for digging so late in, but I was on vacation...

Quoted 8 lines
Quoted 4 lines

Hmm.. the smartweb board has only 4k sram for SPL, and I have no chance to convert it to DM to get rid of some compiler warnings ...

I am unsure what to do now with this hardware ...

And things even get worse: as I wrote in the other thread, after updating to Ubuntu 19.04 as build system, I get gcc 8.3 as cross compiler and smartweb fails to build with that compiler (as the SPL binary is exactly 4k now).

Which reminds me that fixing the various warnings we get with gcc-8.x would be a good general thing to do. :(

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190429/99548995/attachment.sig>
#13 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

In reply to Heiko Schocher
Quoted 66 lines · Heiko Schocher

Hello Simon,

Am 25.04.2019 um 21:24 schrieb Simon Goldschmidt:

Quoted 55 lines

Am 25.04.2019 um 12:50 schrieb Tom Rini:

Quoted 54 lines · Simon Goldschmidt
Quoted 52 lines · Simon Glass
Quoted 38 lines

Hi,

Quoted 25 lines · Tom Rini
Quoted 23 lines · Simon Goldschmidt
Quoted 22 lines · Tom Rini
Quoted 19 lines · Simon Glass
Quoted 7 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

Quoted 8 lines

---

Changes in v4: None
Changes in v3: None

  common/dlmalloc.c | 4 ++++

  1 file changed, 4 insertions(+)
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Ehrm, ok, so how do we proceed?

A good question.  Take a look at spl/u-boot-spl.map on smartweb and see if, of the malloc functions it doesn't discard there's something that maybe could be optimized somewhere?

Quoted 5 lines

I wonder if we should have a Kconfig option like SPL_CHECKS which enables these sorts of minor checks, which may only fix one board at the cost of code size?

Then it could be enabled by default, but disabled on this board?

For a bigger change, this might be an idea, but for a change that I can cut down to 16 or even 8 bytes code size increasement, I don't think having a new option would be good.

Anyway, I just tried at work and I don't get the overflow. Tom, which gcc are you using to get the size error? It works for me on Debian 9 but doesn't work with Ubuntu (both times, default cross compiler toolchain installed).

Quoted 7 lines
Quoted 1 line

I'm using the gcc-7.3 from kernel.org that we use in travis/etc.

Ok, so I have gcc-7.3 on my Ubuntu machine as well. I don't know why 6.3 seems to produce smaller binaries (I thought they were getting smaller with new versions, not larger).

However, I've stripped down that patch to +8 Bytes only and sent v5.

Quoted 8 lines

Thanks!

Sorry for digging so late in, but I was on vacation...

Hmm.. the smartweb board has only 4k sram for SPL, and I have no chance to convert it to DM to get rid of some compiler warnings ...

I am unsure what to do now with this hardware ...

Well, with regards to SPL + DM, this is one of the cases wherein we just have-to allow for the SPL driver code at least to be a one-off. If the "whatever ROM loads of our code" stage can set things up enough such that we can hand off to a full U-Boot, that's great. If not, this is then a case where TPL comes in to play, and that really is as one-off as needed, to load a more general SPL and so forth.

But, I'm fine with saying smartweb keeps and maintains whatever SPL code it needs to use. It's just that in this case, it's not at all a DM thing, it's a change in malloc.

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190429/db4a845c/attachment.sig>
#14 lore raw

PATCH v4 1/2 dlmalloc: fix malloc range at end of ram

Hi,

Quoted 75 lines · Tom Rini
Quoted 73 lines · Heiko Schocher

Hello Simon,

Am 25.04.2019 um 21:24 schrieb Simon Goldschmidt:

Quoted 55 lines

Am 25.04.2019 um 12:50 schrieb Tom Rini:

Quoted 54 lines · Simon Goldschmidt
Quoted 52 lines · Simon Glass
Quoted 38 lines

Hi,

Quoted 25 lines · Tom Rini
Quoted 23 lines · Simon Goldschmidt
Quoted 22 lines · Tom Rini
Quoted 19 lines · Simon Glass
Quoted 7 lines · Simon Goldschmidt

If the malloc range passed to mem_malloc_init() is at the end of address range and 'start + size' overflows to 0, following allocations fail as mem_malloc_end is zero (which looks like uninitialized).

Fix this by subtracting 1 of 'start + size' overflows to zero.

Quoted 8 lines

---

Changes in v4: None
Changes in v3: None

 common/dlmalloc.c | 4 ++++
 1 file changed, 4 insertions(+)
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

So, the problem with this patch is that it increases the generic malloc code size ever so slightly and blows up smartweb :(

Ehrm, ok, so how do we proceed?

A good question. Take a look at spl/u-boot-spl.map on smartweb and see if, of the malloc functions it doesn't discard there's something that maybe could be optimized somewhere?

Quoted 5 lines

I wonder if we should have a Kconfig option like SPL_CHECKS which enables these sorts of minor checks, which may only fix one board at the cost of code size?

Then it could be enabled by default, but disabled on this board?

For a bigger change, this might be an idea, but for a change that I can cut down to 16 or even 8 bytes code size increasement, I don't think having a new option would be good.

Anyway, I just tried at work and I don't get the overflow. Tom, which gcc are you using to get the size error? It works for me on Debian 9 but doesn't work with Ubuntu (both times, default cross compiler toolchain installed).

Quoted 7 lines
Quoted 1 line

I'm using the gcc-7.3 from kernel.org that we use in travis/etc.

Ok, so I have gcc-7.3 on my Ubuntu machine as well. I don't know why 6.3 seems to produce smaller binaries (I thought they were getting smaller with new versions, not larger).

However, I've stripped down that patch to +8 Bytes only and sent v5.

Thanks!

Sorry for digging so late in, but I was on vacation...

Hmm.. the smartweb board has only 4k sram for SPL, and I have no chance to convert it to DM to get rid of some compiler warnings ...

Quoted 8 lines
Quoted 1 line

I am unsure what to do now with this hardware ...

Well, with regards to SPL + DM, this is one of the cases wherein we just have-to allow for the SPL driver code at least to be a one-off. If the "whatever ROM loads of our code" stage can set things up enough such that we can hand off to a full U-Boot, that's great. If not, this is then a case where TPL comes in to play, and that really is as one-off as needed, to load a more general SPL and so forth.

I think DM in SPL is a good thing, so long as there is space. If not, then we should allow non-DM also. I wonder if we need our policy to be written down somewhere?

Quoted 4 lines

But, I'm fine with saying smartweb keeps and maintains whatever SPL code it needs to use. It's just that in this case, it's not at all a DM thing, it's a change in malloc.

Regards,
Simon

#15 lore raw

PATCH v4 2/2 dlmalloc: be compatible to tiny printf

In reply to Simon Goldschmidt

Convert debug output from '%#lx' to '0x%lx' to be compatible with tiny printf used in SPL.

---

Changes in v4:
- dumped clearing BSS before SPL board_init_f, only real bugfixes remain

Changes in v3:
- fixed summary ("stack" -> "heap")
- enable CONFIG_SPL_CLEAR_BSS_F for socfpga_arria10 using full malloc early in SPL
- rebased

 common/dlmalloc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
common/dlmalloc.c +1 -1
index 51d3bd671a..af6f43dcc9 100644
@@ -608,7 +608,7 @@ void mem_malloc_init(ulong start, ulong size)
 		mem_malloc_end--;
 	}
 
-	debug("using memory %#lx-%#lx for malloc()\n", mem_malloc_start,
+	debug("using memory 0x%lx-0x%lx for malloc()\n", mem_malloc_start,
 	      mem_malloc_end);
 #ifdef CONFIG_SYS_MALLOC_CLEAR_ON_INIT
 	memset((void *)mem_malloc_start, 0x0, size);
Signature · 1 line
2.17.1
#16 lore raw

PATCH v4 2/2 dlmalloc: be compatible to tiny printf

Quoted 10 lines · Simon Goldschmidt

Convert debug output from '%#lx' to '0x%lx' to be compatible with tiny printf used in SPL.

---

Changes in v4:
- dumped clearing BSS before SPL board_init_f, only real bugfixes remain

Quoted 8 lines

Changes in v3:
- fixed summary ("stack" -> "heap")
- enable CONFIG_SPL_CLEAR_BSS_F for socfpga_arria10 using full malloc early in SPL
- rebased

 common/dlmalloc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
#17 lore raw

PATCH v4 2/2 dlmalloc: be compatible to tiny printf

In reply to Simon Goldschmidt
Quoted 4 lines · Simon Goldschmidt

Convert debug output from '%#lx' to '0x%lx' to be compatible with tiny printf used in SPL.

Applied to u-boot/master, thanks!

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190424/3dcab9ad/attachment.sig>