* Re: [v11 2/2] misc: pci_endpoint_test: Fix overflow of bar_size
2025-01-21 17:46 ` [v11 2/2] misc: pci_endpoint_test: Fix overflow of bar_size Jon Hunter
@ 2025-01-21 20:23 ` Bjorn Helgaas
2025-01-21 20:31 ` Niklas Cassel
2025-01-22 10:07 ` Niklas Cassel
2 siblings, 0 replies; 8+ messages in thread
From: Bjorn Helgaas @ 2025-01-21 20:23 UTC (permalink / raw)
To: Hans Zhang, Jon Hunter
Cc: manivannan.sadhasivam, kw, kishon, arnd, gregkh, linux-pci,
linux-kernel, rockswang7, Niklas Cassel, linux-tegra
On Tue, Jan 21, 2025 at 05:46:43PM +0000, Jon Hunter wrote:
> On 09/01/2025 09:45, Hans Zhang wrote:
> > With 8GB BAR2, running pcitest -b 2 fails with "TEST FAILED".
> >
> > The return value of the `pci_resource_len` interface is not an integer.
> > Using `pcitest` with an 8GB BAR2, the bar_size of integer type will
> > overflow.
> >
> > Change the data type of bar_size from integer to resource_size_t, to fix
> > the above issue.
> >
> > Signed-off-by: Hans Zhang <18255117159@163.com>
> > Suggested-by: Niklas Cassel <cassel@kernel.org>
> > Reviewed-by: Niklas Cassel <cassel@kernel.org>
> > ---
> > Changes since v10:
> > https://lore.kernel.org/linux-pci/20250108080951.1700230-3-18255117159@163.com/
> >
> > - Replace do_div with the div_u64 API.
> >
> > Changes since v8-v9:
> > https://lore.kernel.org/linux-pci/20250104151652.1652181-1-18255117159@163.com/
> >
> > - Split the patch.
> >
> > Changes since v4-v7:
> > https://lore.kernel.org/linux-pci/20250102120222.1403906-1-18255117159@163.com/
> >
> > - Fix 32-bit OS warnings and errors.
> > - Fix undefined reference to `__udivmoddi4`
> >
> > Changes since v3:
> > https://lore.kernel.org/linux-pci/20241221141009.27317-1-18255117159@163.com/
> >
> > - The patch subject were modified.
> >
> > Changes since v2:
> > https://lore.kernel.org/linux-pci/20241220075253.16791-1-18255117159@163.com/
> >
> > - Fix "changes" part goes below the --- line
> > - The patch commit message were modified.
> >
> > Changes since v1:
> > https://lore.kernel.org/linux-pci/20241217121220.19676-1-18255117159@163.com/
> >
> > - The patch subject and commit message were modified.
> > ---
> > drivers/misc/pci_endpoint_test.c | 5 +++--
> > 1 file changed, 3 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/pci_endpoint_test.c
> > index f78c7540c52c..0f6291801078 100644
> > --- a/drivers/misc/pci_endpoint_test.c
> > +++ b/drivers/misc/pci_endpoint_test.c
> > @@ -280,10 +280,11 @@ static int pci_endpoint_test_bar_memcmp(struct pci_endpoint_test *test,
> > static bool pci_endpoint_test_bar(struct pci_endpoint_test *test,
> > enum pci_barno barno)
> > {
> > - int j, bar_size, buf_size, iters;
> > + int j, buf_size, iters;
> > void *write_buf __free(kfree) = NULL;
> > void *read_buf __free(kfree) = NULL;
> > struct pci_dev *pdev = test->pdev;
> > + resource_size_t bar_size;
> > if (!test->bar[barno])
> > return false;
> > @@ -307,7 +308,7 @@ static bool pci_endpoint_test_bar(struct pci_endpoint_test *test,
> > if (!read_buf)
> > return false;
> > - iters = bar_size / buf_size;
> > + iters = div_u64(bar_size, buf_size);
> > for (j = 0; j < iters; j++)
> > if (pci_endpoint_test_bar_memcmp(test, barno, buf_size * j,
> > write_buf, read_buf, buf_size))
>
>
> This change breaks building the kernel with GCC v7 and I see ...
>
> ERROR: modpost: "__aeabi_uldivmod" [drivers/misc/pci_endpoint_test.ko]
> undefined!
> ERROR: modpost: "__aeabi_ldivmod" [drivers/misc/pci_endpoint_test.ko]
> undefined!
>
> I know that this is an old GCC version, but this is a farm builder and the
> kernel still indicates that GCC v5.1 is still supported [0].
>
> Jon
>
> [0] https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/Documentation/process/changes.rst
Dropping from pci/endpoint-test for now until this gets sorted out.
For reference, the current iteration (the one I'm dropping) is below:
commit c6e73776c0c4 ("misc: pci_endpoint_test: Prevent bar_size overflow")
Author: Hans Zhang <18255117159@163.com>
Date: Thu Jan 9 17:45:56 2025 +0800
misc: pci_endpoint_test: Prevent bar_size overflow
Running 'pcitest -b 2' fails with "TEST FAILED" when the BAR2 size
exceeds 4GB.
The return value of the pci_resource_len() macro can be larger than that
of a signed integer type. As such, for example, when using 'pcitest'
with an 8GB BAR2, the bar_size of the integer type will overflow.
Change bar_size from integer to resource_size_t to prevent integer
overflow for large BAR sizes with 32-bit compilers. Similarly, use
div_u64() to perform 64-bit division in that situation.
Link: https://lore.kernel.org/r/20250109094556.1724663-3-18255117159@163.com
Signed-off-by: Hans Zhang <18255117159@163.com>
[kwilczynski: commit log]
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
[bhelgaas: pick up commit log update from
https://lore.kernel.org/r/Z4jTEkznMUcApzbe@ryzen]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/pci_endpoint_test.c
index be0c7c870e6a..abe6c8f03fca 100644
--- a/drivers/misc/pci_endpoint_test.c
+++ b/drivers/misc/pci_endpoint_test.c
@@ -283,10 +283,11 @@ static int pci_endpoint_test_bar_memcmp(struct pci_endpoint_test *test,
static bool pci_endpoint_test_bar(struct pci_endpoint_test *test,
enum pci_barno barno)
{
- int j, bar_size, buf_size, iters;
+ int j, buf_size, iters;
void *write_buf __free(kfree) = NULL;
void *read_buf __free(kfree) = NULL;
struct pci_dev *pdev = test->pdev;
+ resource_size_t bar_size;
if (!test->bar[barno])
return false;
@@ -310,7 +311,7 @@ static bool pci_endpoint_test_bar(struct pci_endpoint_test *test,
if (!read_buf)
return false;
- iters = bar_size / buf_size;
+ iters = div_u64(bar_size, buf_size);
for (j = 0; j < iters; j++)
if (pci_endpoint_test_bar_memcmp(test, barno, buf_size * j,
write_buf, read_buf, buf_size))
^ permalink raw reply related [flat|nested] 8+ messages in thread* Re: [v11 2/2] misc: pci_endpoint_test: Fix overflow of bar_size
2025-01-21 17:46 ` [v11 2/2] misc: pci_endpoint_test: Fix overflow of bar_size Jon Hunter
2025-01-21 20:23 ` Bjorn Helgaas
@ 2025-01-21 20:31 ` Niklas Cassel
2025-01-22 6:03 ` Jon Hunter
2025-01-22 10:07 ` Niklas Cassel
2 siblings, 1 reply; 8+ messages in thread
From: Niklas Cassel @ 2025-01-21 20:31 UTC (permalink / raw)
To: Jon Hunter, Hans Zhang, manivannan.sadhasivam
Cc: kw, kishon, arnd, gregkh, linux-pci, linux-kernel, rockswang7,
linux-tegra@vger.kernel.org
On 21 January 2025 18:46:43 CET, Jon Hunter <jonathanh@nvidia.com> wrote:
>Hi Hans,
>
>On 09/01/2025 09:45, Hans Zhang wrote:
>> With 8GB BAR2, running pcitest -b 2 fails with "TEST FAILED".
>>
>> The return value of the `pci_resource_len` interface is not an integer.
>> Using `pcitest` with an 8GB BAR2, the bar_size of integer type will
>> overflow.
>>
>> Change the data type of bar_size from integer to resource_size_t, to fix
>> the above issue.
>>
>> Signed-off-by: Hans Zhang <18255117159@163.com>
>> Suggested-by: Niklas Cassel <cassel@kernel.org>
>> Reviewed-by: Niklas Cassel <cassel@kernel.org>
>> ---
>> Changes since v10:
>> https://lore.kernel.org/linux-pci/20250108080951.1700230-3-18255117159@163.com/
>>
>> - Replace do_div with the div_u64 API.
>>
>> Changes since v8-v9:
>> https://lore.kernel.org/linux-pci/20250104151652.1652181-1-18255117159@163.com/
>>
>> - Split the patch.
>>
>> Changes since v4-v7:
>> https://lore.kernel.org/linux-pci/20250102120222.1403906-1-18255117159@163.com/
>>
>> - Fix 32-bit OS warnings and errors.
>> - Fix undefined reference to `__udivmoddi4`
>>
>> Changes since v3:
>> https://lore.kernel.org/linux-pci/20241221141009.27317-1-18255117159@163.com/
>>
>> - The patch subject were modified.
>>
>> Changes since v2:
>> https://lore.kernel.org/linux-pci/20241220075253.16791-1-18255117159@163.com/
>>
>> - Fix "changes" part goes below the --- line
>> - The patch commit message were modified.
>>
>> Changes since v1:
>> https://lore.kernel.org/linux-pci/20241217121220.19676-1-18255117159@163.com/
>>
>> - The patch subject and commit message were modified.
>> ---
>> drivers/misc/pci_endpoint_test.c | 5 +++--
>> 1 file changed, 3 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/pci_endpoint_test.c
>> index f78c7540c52c..0f6291801078 100644
>> --- a/drivers/misc/pci_endpoint_test.c
>> +++ b/drivers/misc/pci_endpoint_test.c
>> @@ -280,10 +280,11 @@ static int pci_endpoint_test_bar_memcmp(struct pci_endpoint_test *test,
>> static bool pci_endpoint_test_bar(struct pci_endpoint_test *test,
>> enum pci_barno barno)
>> {
>> - int j, bar_size, buf_size, iters;
>> + int j, buf_size, iters;
>> void *write_buf __free(kfree) = NULL;
>> void *read_buf __free(kfree) = NULL;
>> struct pci_dev *pdev = test->pdev;
>> + resource_size_t bar_size;
>> if (!test->bar[barno])
>> return false;
>> @@ -307,7 +308,7 @@ static bool pci_endpoint_test_bar(struct pci_endpoint_test *test,
>> if (!read_buf)
>> return false;
>> - iters = bar_size / buf_size;
>> + iters = div_u64(bar_size, buf_size);
>> for (j = 0; j < iters; j++)
>> if (pci_endpoint_test_bar_memcmp(test, barno, buf_size * j,
>> write_buf, read_buf, buf_size))
>
>
>This change breaks building the kernel with GCC v7 and I see ...
>
>ERROR: modpost: "__aeabi_uldivmod" [drivers/misc/pci_endpoint_test.ko] undefined!
>ERROR: modpost: "__aeabi_ldivmod" [drivers/misc/pci_endpoint_test.ko] undefined!
>
>I know that this is an old GCC version, but this is a farm builder and the kernel still indicates that GCC v5.1 is still supported [0].
>
Hmm.. div_u64() should work on gcc 5.1.
Out of curiosity, is there any difference if you add:
#include <linux/math64.h>
to
drivers/misc/pci_endpoint_test.c ?
Kind regards,
Niklas
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [v11 2/2] misc: pci_endpoint_test: Fix overflow of bar_size
2025-01-21 20:31 ` Niklas Cassel
@ 2025-01-22 6:03 ` Jon Hunter
0 siblings, 0 replies; 8+ messages in thread
From: Jon Hunter @ 2025-01-22 6:03 UTC (permalink / raw)
To: Niklas Cassel, Hans Zhang, manivannan.sadhasivam
Cc: kw, kishon, arnd, gregkh, linux-pci, linux-kernel, rockswang7,
linux-tegra@vger.kernel.org
On 21/01/2025 20:31, Niklas Cassel wrote:
>
>
> On 21 January 2025 18:46:43 CET, Jon Hunter <jonathanh@nvidia.com> wrote:
>> Hi Hans,
>>
>> On 09/01/2025 09:45, Hans Zhang wrote:
>>> With 8GB BAR2, running pcitest -b 2 fails with "TEST FAILED".
>>>
>>> The return value of the `pci_resource_len` interface is not an integer.
>>> Using `pcitest` with an 8GB BAR2, the bar_size of integer type will
>>> overflow.
>>>
>>> Change the data type of bar_size from integer to resource_size_t, to fix
>>> the above issue.
>>>
>>> Signed-off-by: Hans Zhang <18255117159@163.com>
>>> Suggested-by: Niklas Cassel <cassel@kernel.org>
>>> Reviewed-by: Niklas Cassel <cassel@kernel.org>
>>> ---
>>> Changes since v10:
>>> https://lore.kernel.org/linux-pci/20250108080951.1700230-3-18255117159@163.com/
>>>
>>> - Replace do_div with the div_u64 API.
>>>
>>> Changes since v8-v9:
>>> https://lore.kernel.org/linux-pci/20250104151652.1652181-1-18255117159@163.com/
>>>
>>> - Split the patch.
>>>
>>> Changes since v4-v7:
>>> https://lore.kernel.org/linux-pci/20250102120222.1403906-1-18255117159@163.com/
>>>
>>> - Fix 32-bit OS warnings and errors.
>>> - Fix undefined reference to `__udivmoddi4`
>>>
>>> Changes since v3:
>>> https://lore.kernel.org/linux-pci/20241221141009.27317-1-18255117159@163.com/
>>>
>>> - The patch subject were modified.
>>>
>>> Changes since v2:
>>> https://lore.kernel.org/linux-pci/20241220075253.16791-1-18255117159@163.com/
>>>
>>> - Fix "changes" part goes below the --- line
>>> - The patch commit message were modified.
>>>
>>> Changes since v1:
>>> https://lore.kernel.org/linux-pci/20241217121220.19676-1-18255117159@163.com/
>>>
>>> - The patch subject and commit message were modified.
>>> ---
>>> drivers/misc/pci_endpoint_test.c | 5 +++--
>>> 1 file changed, 3 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/pci_endpoint_test.c
>>> index f78c7540c52c..0f6291801078 100644
>>> --- a/drivers/misc/pci_endpoint_test.c
>>> +++ b/drivers/misc/pci_endpoint_test.c
>>> @@ -280,10 +280,11 @@ static int pci_endpoint_test_bar_memcmp(struct pci_endpoint_test *test,
>>> static bool pci_endpoint_test_bar(struct pci_endpoint_test *test,
>>> enum pci_barno barno)
>>> {
>>> - int j, bar_size, buf_size, iters;
>>> + int j, buf_size, iters;
>>> void *write_buf __free(kfree) = NULL;
>>> void *read_buf __free(kfree) = NULL;
>>> struct pci_dev *pdev = test->pdev;
>>> + resource_size_t bar_size;
>>> if (!test->bar[barno])
>>> return false;
>>> @@ -307,7 +308,7 @@ static bool pci_endpoint_test_bar(struct pci_endpoint_test *test,
>>> if (!read_buf)
>>> return false;
>>> - iters = bar_size / buf_size;
>>> + iters = div_u64(bar_size, buf_size);
>>> for (j = 0; j < iters; j++)
>>> if (pci_endpoint_test_bar_memcmp(test, barno, buf_size * j,
>>> write_buf, read_buf, buf_size))
>>
>>
>> This change breaks building the kernel with GCC v7 and I see ...
>>
>> ERROR: modpost: "__aeabi_uldivmod" [drivers/misc/pci_endpoint_test.ko] undefined!
>> ERROR: modpost: "__aeabi_ldivmod" [drivers/misc/pci_endpoint_test.ko] undefined!
>>
>> I know that this is an old GCC version, but this is a farm builder and the kernel still indicates that GCC v5.1 is still supported [0].
>>
>
> Hmm.. div_u64() should work on gcc 5.1.
>
> Out of curiosity, is there any difference if you add:
>
> #include <linux/math64.h>
> to
> drivers/misc/pci_endpoint_test.c ?
I tried the above, but I am still seeing the same problem. So
unfortunately, that did not fix it.
Cheers
Jon
--
nvpublic
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [v11 2/2] misc: pci_endpoint_test: Fix overflow of bar_size
2025-01-21 17:46 ` [v11 2/2] misc: pci_endpoint_test: Fix overflow of bar_size Jon Hunter
2025-01-21 20:23 ` Bjorn Helgaas
2025-01-21 20:31 ` Niklas Cassel
@ 2025-01-22 10:07 ` Niklas Cassel
2025-01-22 10:26 ` Arnd Bergmann
2 siblings, 1 reply; 8+ messages in thread
From: Niklas Cassel @ 2025-01-22 10:07 UTC (permalink / raw)
To: Arnd Bergmann
Cc: Hans Zhang, manivannan.sadhasivam, kw, kishon, gregkh, linux-pci,
linux-kernel, rockswang7, linux-tegra@vger.kernel.org, Jon Hunter
On Tue, Jan 21, 2025 at 05:46:43PM +0000, Jon Hunter wrote:
> > diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/pci_endpoint_test.c
> > index f78c7540c52c..0f6291801078 100644
> > --- a/drivers/misc/pci_endpoint_test.c
> > +++ b/drivers/misc/pci_endpoint_test.c
> > @@ -280,10 +280,11 @@ static int pci_endpoint_test_bar_memcmp(struct pci_endpoint_test *test,
> > static bool pci_endpoint_test_bar(struct pci_endpoint_test *test,
> > enum pci_barno barno)
> > {
> > - int j, bar_size, buf_size, iters;
> > + int j, buf_size, iters;
> > void *write_buf __free(kfree) = NULL;
> > void *read_buf __free(kfree) = NULL;
> > struct pci_dev *pdev = test->pdev;
> > + resource_size_t bar_size;
> > if (!test->bar[barno])
> > return false;
> > @@ -307,7 +308,7 @@ static bool pci_endpoint_test_bar(struct pci_endpoint_test *test,
> > if (!read_buf)
> > return false;
> > - iters = bar_size / buf_size;
> > + iters = div_u64(bar_size, buf_size);
> > for (j = 0; j < iters; j++)
> > if (pci_endpoint_test_bar_memcmp(test, barno, buf_size * j,
> > write_buf, read_buf, buf_size))
>
>
> This change breaks building the kernel with GCC v7 and I see ...
>
> ERROR: modpost: "__aeabi_uldivmod" [drivers/misc/pci_endpoint_test.ko]
> undefined!
> ERROR: modpost: "__aeabi_ldivmod" [drivers/misc/pci_endpoint_test.ko]
> undefined!
>
> I know that this is an old GCC version, but this is a farm builder and the
> kernel still indicates that GCC v5.1 is still supported [0].
Arnd,
do you have any idea what is going on here?
I'm a bit puzzled, since looking at other reports of this error,
e.g.:
https://lore.kernel.org/all/20241018151016.3496613-1-arnd@kernel.org/
using div_u64() is usually the solution for this problem (for things that
are not performance critical), not the cause of it... any ideas?
Kind regards,
Niklas
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [v11 2/2] misc: pci_endpoint_test: Fix overflow of bar_size
2025-01-22 10:07 ` Niklas Cassel
@ 2025-01-22 10:26 ` Arnd Bergmann
2025-01-22 11:18 ` Jon Hunter
0 siblings, 1 reply; 8+ messages in thread
From: Arnd Bergmann @ 2025-01-22 10:26 UTC (permalink / raw)
To: Niklas Cassel
Cc: Hans Zhang, Manivannan Sadhasivam, Krzysztof Wilczyński,
Kishon Vijay Abraham I, Greg Kroah-Hartman, linux-pci,
linux-kernel, rockswang7, linux-tegra@vger.kernel.org, Jon Hunter
On Wed, Jan 22, 2025, at 11:07, Niklas Cassel wrote:
> On Tue, Jan 21, 2025 at 05:46:43PM +0000, Jon Hunter wrote:
>> > diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/>>
>>
>> This change breaks building the kernel with GCC v7 and I see ...
>>
>> ERROR: modpost: "__aeabi_uldivmod" [drivers/misc/pci_endpoint_test.ko]
>> undefined!
>> ERROR: modpost: "__aeabi_ldivmod" [drivers/misc/pci_endpoint_test.ko]
>> undefined!
>>
>> I know that this is an old GCC version, but this is a farm builder and the
>> kernel still indicates that GCC v5.1 is still supported [0].
>
> do you have any idea what is going on here?
>
> I'm a bit puzzled, since looking at other reports of this error,
> e.g.:
> https://lore.kernel.org/all/20241018151016.3496613-1-arnd@kernel.org/
>
> using div_u64() is usually the solution for this problem (for things that
> are not performance critical), not the cause of it... any ideas?
I have tried to look at the email thread, but not tried to reproduce
it yet. I have two ideas about what might be happening:
a) something causes a /different/ division to call into
__aeabi_uldivmod(), not the one from div_u64().
b) the compiler notices one of the arguments to div_u64() being
constant in some cases and splits the calling function into
two special cases, for both the constant and non-constant
cases. This sometimes confuses the __builtin_constant_p()
in do_div() that decides to fall back to a 32-bit division.
Try looking at the .s file when you run
'make drivers/misc/pci_endpoint_test.s' to see where exactly it
calls the two division functions, maybe you see the problem then,
otherwise I can try to reproduce it here.
Arnd
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [v11 2/2] misc: pci_endpoint_test: Fix overflow of bar_size
2025-01-22 10:26 ` Arnd Bergmann
@ 2025-01-22 11:18 ` Jon Hunter
2025-01-23 8:10 ` Niklas Cassel
0 siblings, 1 reply; 8+ messages in thread
From: Jon Hunter @ 2025-01-22 11:18 UTC (permalink / raw)
To: Arnd Bergmann, Niklas Cassel
Cc: Hans Zhang, Manivannan Sadhasivam, Krzysztof Wilczyński,
Kishon Vijay Abraham I, Greg Kroah-Hartman, linux-pci,
linux-kernel, rockswang7, linux-tegra@vger.kernel.org
On 22/01/2025 10:26, Arnd Bergmann wrote:
> On Wed, Jan 22, 2025, at 11:07, Niklas Cassel wrote:
>> On Tue, Jan 21, 2025 at 05:46:43PM +0000, Jon Hunter wrote:
>>>> diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/>>
>>>
>>> This change breaks building the kernel with GCC v7 and I see ...
>>>
>>> ERROR: modpost: "__aeabi_uldivmod" [drivers/misc/pci_endpoint_test.ko]
>>> undefined!
>>> ERROR: modpost: "__aeabi_ldivmod" [drivers/misc/pci_endpoint_test.ko]
>>> undefined!
>>>
>>> I know that this is an old GCC version, but this is a farm builder and the
>>> kernel still indicates that GCC v5.1 is still supported [0].
>>
>> do you have any idea what is going on here?
>>
>> I'm a bit puzzled, since looking at other reports of this error,
>> e.g.:
>> https://lore.kernel.org/all/20241018151016.3496613-1-arnd@kernel.org/
>>
>> using div_u64() is usually the solution for this problem (for things that
>> are not performance critical), not the cause of it... any ideas?
>
> I have tried to look at the email thread, but not tried to reproduce
> it yet. I have two ideas about what might be happening:
>
> a) something causes a /different/ division to call into
> __aeabi_uldivmod(), not the one from div_u64().
>
> b) the compiler notices one of the arguments to div_u64() being
> constant in some cases and splits the calling function into
> two special cases, for both the constant and non-constant
> cases. This sometimes confuses the __builtin_constant_p()
> in do_div() that decides to fall back to a 32-bit division.
>
> Try looking at the .s file when you run
> 'make drivers/misc/pci_endpoint_test.s' to see where exactly it
> calls the two division functions, maybe you see the problem then,
> otherwise I can try to reproduce it here.
FWIW I have encountered similar problems with this before with this
compiler ...
https://lore.kernel.org/linux-tegra/7733a4ca-330b-4127-af12-33f376fbbc47@nvidia.com/
Jon
--
nvpublic
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [v11 2/2] misc: pci_endpoint_test: Fix overflow of bar_size
2025-01-22 11:18 ` Jon Hunter
@ 2025-01-23 8:10 ` Niklas Cassel
0 siblings, 0 replies; 8+ messages in thread
From: Niklas Cassel @ 2025-01-23 8:10 UTC (permalink / raw)
To: Jon Hunter
Cc: Arnd Bergmann, Hans Zhang, Manivannan Sadhasivam,
Krzysztof Wilczyński, Kishon Vijay Abraham I,
Greg Kroah-Hartman, linux-pci, linux-kernel, rockswang7,
linux-tegra@vger.kernel.org
On Wed, Jan 22, 2025 at 11:18:44AM +0000, Jon Hunter wrote:
>
> On 22/01/2025 10:26, Arnd Bergmann wrote:
> > On Wed, Jan 22, 2025, at 11:07, Niklas Cassel wrote:
> > > On Tue, Jan 21, 2025 at 05:46:43PM +0000, Jon Hunter wrote:
> > > > > diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/>>
> > > >
> > > > This change breaks building the kernel with GCC v7 and I see ...
> > > >
> > > > ERROR: modpost: "__aeabi_uldivmod" [drivers/misc/pci_endpoint_test.ko]
> > > > undefined!
> > > > ERROR: modpost: "__aeabi_ldivmod" [drivers/misc/pci_endpoint_test.ko]
> > > > undefined!
> > > >
> > > > I know that this is an old GCC version, but this is a farm builder and the
> > > > kernel still indicates that GCC v5.1 is still supported [0].
> > >
> > > do you have any idea what is going on here?
> > >
> > > I'm a bit puzzled, since looking at other reports of this error,
> > > e.g.:
> > > https://lore.kernel.org/all/20241018151016.3496613-1-arnd@kernel.org/
> > >
> > > using div_u64() is usually the solution for this problem (for things that
> > > are not performance critical), not the cause of it... any ideas?
> >
> > I have tried to look at the email thread, but not tried to reproduce
> > it yet. I have two ideas about what might be happening:
> >
> > a) something causes a /different/ division to call into
> > __aeabi_uldivmod(), not the one from div_u64().
> >
> > b) the compiler notices one of the arguments to div_u64() being
> > constant in some cases and splits the calling function into
> > two special cases, for both the constant and non-constant
> > cases. This sometimes confuses the __builtin_constant_p()
> > in do_div() that decides to fall back to a 32-bit division.
> >
> > Try looking at the .s file when you run
> > 'make drivers/misc/pci_endpoint_test.s' to see where exactly it
> > calls the two division functions, maybe you see the problem then,
> > otherwise I can try to reproduce it here.
>
>
> FWIW I have encountered similar problems with this before with this compiler
> ...
>
> https://lore.kernel.org/linux-tegra/7733a4ca-330b-4127-af12-33f376fbbc47@nvidia.com/
I had no errors when building this patch using:
gcc (Debian 7.4.0-6) 7.4.0
on i386.
neither with CONFIG_PHYS_ADDR_T_64BIT=y (CONFIG_X86_PAE=y)
nor without it set.
So perhaps this is a bug in gcc 7.x on arm32 or in the arm32
implementation of div_u64() ?
Kind regards,
Niklas
^ permalink raw reply [flat|nested] 8+ messages in thread