PCI/ASPM: Use 64-bit arithmetic instead of 32-bit
PATCH 1/1 PCI/ASPM: Use 64-bit arithmetic instead of 32-bit
Add suffix ULL to constant 1000 in order to give the compiler complete information about the proper arithmetic to use. Notice that this constant is used in a context that expects an expression of type u64 (64 bits, unsigned).
The expression threshold_us * 1000 is currently being evaluated using 32-bit arithmetic.
Addresses-Coverity-ID: 1462501
--- drivers/pci/pcie/aspm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/pci/pcie/aspm.c +1 -1
@@ -322,7 +322,7 @@ static u32 calc_l1ss_pwron(struct pci_dev *pdev, u32 scale, u32 val)static void encode_l12_threshold (u32 threshold_us,u32 *scale,u32 *value) { -u64 threshold_ns = threshold_us *1000 ; +u64 threshold_ns = threshold_us *1000ULL ;/* See PCIe r3.1, sec 7.33.3 and sec 6.18 */ if (threshold_ns <32 ) {
Signature · 1 line
2.7.4
drivers/pci/pcie/aspm.c static u32 calc_l1ss_pwron(struct pci_dev *pdev, u32 scale, u32 val) · line 325
Quoted 10 lines · Gustavo A. R. Silva
Add suffix ULL to constant 1000 in order to give the compiler complete information about the proper arithmetic to use. Notice that this constant is used in a context that expects an expression of type u64 (64 bits, unsigned).
The expression
threshold_us* 1000 is currently being evaluated using 32-bit arithmetic.drivers/pci/pcie/aspm.c +1
-1-u64 threshold_ns = threshold_us *1000 ; +u64 threshold_ns = threshold_us *1000ULL ;
Shouldn't be other way around, i.e.
(u64)threshold_us ?
But still the question. have you checked all callers? Does it even makes sense?
Signature · 2 lines
With Best Regards, Andy Shevchenko
Hi Andy,
Quoting Andy Shevchenko <[email protected]>:
Quoted 10 lines
Quoted 7 lines · Gustavo A. R. Silva
Add suffix ULL to constant 1000 in order to give the compiler complete information about the proper arithmetic to use. Notice that this constant is used in a context that expects an expression of type u64 (64 bits, unsigned).
The expression
threshold_us* 1000 is currently being evaluated using 32-bit arithmetic.
Quoted 7 lines
drivers/pci/pcie/aspm.c static u32 calc_l1ss_pwron(struct pci_dev *pdev, u32 scale, u32 val) · line 325
Quoted 2 lines
drivers/pci/pcie/aspm.c +1
-1-u64 threshold_ns = threshold_us *1000 ; +u64 threshold_ns = threshold_us *1000ULL ;Shouldn't be other way around, i.e.
(u64)threshold_us ?
Either way works. The thing is that casting threshold_us to u64 may imply that there is something wrong with threshold_us, which does not seem to be the case. So adding the suffix ULL to the constant 1000 is good enough to make the expression be evaluated using 64-bit arithmetic instead of 32-bit.
But, again, either way works.
Quoted 2 lines
But still the question. have you checked all callers? Does it even makes sense?
The proposed patch was due to fact that currently threshold_ns is of type u64. But based on the following piece of code (which is the only piece of code from where encode_l12_threshold is being called):
* Based on PCIe r3.1, sec 5.5.3.3.1, Figures 5-16 and 5-17, and
* Table 5-11. T(POWER_OFF) is at most 2us and T(L1.2) is at
* least 4us.
*/
l1_2_threshold = 2 + 4 + t_common_mode + t_power_on; encode_l12_threshold(l1_2_threshold, &scale, &value);
It seems to me that it makes no sense for threshold_ns to be of type u64, because the expression threshold_us * 1000 will never exceed the 32-bit limits. So if you agree I can send a patch to change its type to u32 instead.
Thanks
--
Gustavo
Quoted 4 lines · Gustavo A. R. Silva
It seems to me that it makes no sense for
threshold_nsto be of type u64, because the expressionthreshold_us* 1000 will never exceed the 32-bit limits. So if you agree I can send a patch to change its type to u32 instead.
Whatever Bjorn prefers.
Signature · 2 lines
With Best Regards, Andy Shevchenko
Quoted 44 lines · Gustavo A. R. Silva
Hi Andy,
Quoting Andy Shevchenko <[email protected]>:
Quoted 10 lines
Quoted 7 lines · Gustavo A. R. Silva
Add suffix ULL to constant 1000 in order to give the compiler complete information about the proper arithmetic to use. Notice that this constant is used in a context that expects an expression of type u64 (64 bits, unsigned).
The expression
threshold_us* 1000 is currently being evaluated using 32-bit arithmetic.Quoted 7 lines
drivers/pci/pcie/aspm.c static u32 calc_l1ss_pwron(struct pci_dev *pdev, u32 scale, u32 val) · line 325
Quoted 2 lines
drivers/pci/pcie/aspm.c +1
-1-u64 threshold_ns = threshold_us *1000 ; +u64 threshold_ns = threshold_us *1000ULL ;Shouldn't be other way around, i.e.
(u64)threshold_us ?
Either way works. The thing is that casting
threshold_usto u64 may imply that there is something wrong withthreshold_us, which does not seem to be the case. So adding the suffix ULL to the constant 1000 is good enough to make the expression be evaluated using 64-bit arithmetic instead of 32-bit.But, again, either way works.
Quoted 3 lines
But still the question. have you checked all callers? Does it even makes sense?
The proposed patch was due to fact that currently
threshold_nsis of type u64. But based on the following piece of code (which is the only piece of code from whereencode_l12_thresholdis being called):* Based on PCIe r3.1, sec 5.5.3.3.1, Figures 5-16 and 5-17, and
* Table 5-11.T(POWER_OFF)is at most 2us and T(L1.2) is at
* least 4us.
*/
l1_2_threshold= 2 + 4 +t_common_mode+t_power_on;encode_l12_threshold(l1_2_threshold, &scale, &value);
Quoted 4 lines
It seems to me that it makes no sense for
threshold_nsto be of type u64, because the expressionthreshold_us* 1000 will never exceed the 32-bit limits. So if you agree I can send a patch to change its type to u32 instead.
Changing it to u32 sounds good to me. I can't remember why I chose u64 to begin with, but it doesn't look necessary.
Thanks for cleaning this up!