[PATCH] pcie: add pcie aspm registers check on root port and device.

Alex Hung alex.hung at canonical.com
Fri Feb 10 07:14:10 UTC 2012


On 02/09/2012 07:31 PM, Colin Ian King wrote:
> On 08/02/12 09:29, Alex Hung wrote:
>> Signed-off-by: Alex Hung<alex.hung at canonical.com>
>> ---
>>   src/pci/aspm/aspm.c |  219 
>> ++++++++++++++++++++++++++++++++++++++++++++++++++-
>>   1 files changed, 216 insertions(+), 3 deletions(-)
>>
>> diff --git a/src/pci/aspm/aspm.c b/src/pci/aspm/aspm.c
>> index db82718..9ed7fc0 100644
>> --- a/src/pci/aspm/aspm.c
>> +++ b/src/pci/aspm/aspm.c
>> @@ -26,10 +26,45 @@
>>   #include<string.h>
>>   #include<fcntl.h>
>>   #include<limits.h>
>> -#include<linux/pci.h>
>> -
>>   #include "fwts.h"
>>
>> +struct pci_device {
>> +    uint8_t segment;
>> +    uint8_t bus;
>> +    uint8_t dev;
>> +    uint8_t func;
>> +    uint8_t config[256];
>> +    struct pci_device *next;
>> +};
>> +
>> +
>> +struct pcie_capability {
>> +    uint8_t pcie_cap_id;
>> +    uint8_t next_cap_point;
>> +    uint16_t pcie_cap_reg;
>> +    uint32_t device_cap;
>> +    uint16_t device_contrl;
>> +    uint16_t device_status;
>> +    uint32_t link_cap;
>> +    uint16_t link_contrl;
>> +    uint16_t link_status;
>> +    uint32_t slot_cap;
>> +    uint16_t slot_contrl;
>> +    uint16_t slot_status;
>> +    uint16_t root_contrl;
>> +    uint16_t root_cap;
>> +    uint32_t root_status;
>> +    uint32_t device_cap2;
>> +    uint16_t device_contrl2;
>> +    uint16_t device_status2;
>> +    uint32_t link_cap2;
>> +    uint16_t link_contrl2;
>> +    uint16_t link_status2;
>> +    uint32_t slot_cap2;
>> +    uint16_t slot_contrl2;
>> +    uint16_t slot_status2;
>> +};
>> +
>
> Can you put a comment in referencing which specification I can find a 
> description of these data structures?
>
>>   static int facp_get_aspm_control(fwts_framework *fw, int *aspm)
>>   {
>>       fwts_acpi_table_info *table;
>> @@ -51,6 +86,174 @@ static int facp_get_aspm_control(fwts_framework 
>> *fw, int *aspm)
>>       return FWTS_OK;
>>   }
>>
>> +int pcie_compare_rp_dev_aspm_registers(fwts_framework *fw,
>> +                        struct pci_device *rp,
>> +                        struct pci_device *dev)
>> +{
>> +    struct pcie_capability *rp_cap, *device_cap;
>> +    uint8_t rp_aspm_support, rp_aspm_cntrl;
>> +    uint8_t device_aspm_support, device_aspm_cntrl;
>> +    uint8_t next_cap;
>> +    int ret = FWTS_OK;
>> +
>> +    next_cap = rp->config[0x34];
>
> #define of the 0x34 would be useful as I'm lazy and don't want to look 
> it up in the PCI spec :-)
>
>> +    rp_cap = (struct pcie_capability *)&rp->config[next_cap];
>> +    while (rp_cap->pcie_cap_id != 0x10) {
>
> same here for 0x10 - what is that?
>
>> +        if (rp_cap->next_cap_point == 0x00)
>> +            break;
>> +        next_cap = rp_cap->next_cap_point;
>> +        rp_cap = (struct pcie_capability *)&rp->config[next_cap];
>> +    }
>> +
>> +    next_cap = dev->config[0x34];
>> +    device_cap = (struct pcie_capability *) &dev->config[next_cap];
>> +    while (device_cap->pcie_cap_id != 0x10) {
>> +        if (device_cap->next_cap_point == 0x00)
>> +            break;
>> +        next_cap = device_cap->next_cap_point;
>> +        device_cap = (struct pcie_capability *) &dev->config[next_cap];
>> +    }
>> +
>> +    rp_aspm_support = (rp_cap->link_cap&  0x0c00)>>  10;
>> +    rp_aspm_cntrl = (rp_cap->link_contrl&  0x03);
>> +    device_aspm_support = (device_cap->link_cap&  0x0c00)>>  10;
>> +    device_aspm_cntrl = (device_cap->link_contrl&  0x03);
>> +
>> +    if ((rp_aspm_support&  0x01) != (rp_aspm_cntrl&  0x01)) {
>> +        fwts_warning(fw, "RP %02Xh:%02Xh.%02Xh L0s not enabled.",
>> +             rp->bus, rp->dev, rp->func);
>> +    }
>> +
>> +    if ((rp_aspm_support&  0x02) != (rp_aspm_cntrl&  0x02)) {
>> +        fwts_warning(fw, "RP %02Xh:%02Xh.%02Xh L1 not enabled.",
>> +            rp->bus, rp->dev, rp->func);
>> +    }
>> +
> bunch more #defines on the magic values would be handy too..
>> +
>> +    if ((device_aspm_support&  0x01) != (device_aspm_cntrl&  0x01)) {
>> +        fwts_warning(fw, "Device %02Xh:%02Xh.%02Xh L0s not enabled.",
>> +            dev->bus, dev->dev, dev->func);
>> +    }
>> +
>> +    if ((device_aspm_support&  0x02) != (device_aspm_cntrl&  0x02)) {
>> +        fwts_warning(fw, "Device %02Xh:%02Xh.%02Xh L1 not enabled.",
>> +            dev->bus, dev->dev, dev->func);
>> +    }
>> +
>> +    if (rp_aspm_cntrl != device_aspm_cntrl) {
>> +        fwts_failed(fw, LOG_LEVEL_MEDIUM, "PCIEASPM_UNMATCHED",
>> +            "PCIE aspm setting was not matched.");
>> +        fwts_log_error(fw, "RP %02Xh:%02Xh.%02Xh has aspm = %02Xh."
>> +            , rp->bus, rp->dev, rp->func, rp_aspm_cntrl);
>> +        fwts_log_error(fw, "Device %02Xh:%02Xh.%02Xh has aspm = 
>> %02Xh.",
>> +            dev->bus, dev->dev, dev->func, device_aspm_cntrl);
>> +    } else {
>> +        fwts_passed(fw, "PCIE aspm setting matched was matched.");
>> +    }
>> +
>> +    return ret;
>> +}
>> +
>> +int pcie_check_aspm_registers(fwts_framework *fw,
>> +                   const fwts_log_level level)
>> +{
>> +    fwts_list *lspci_output;
>> +    fwts_list_link *item;
>> +    struct pci_device *root = NULL, *cur = NULL, *device = NULL;
>> +    char command[PATH_MAX];
>> +    int ret = FWTS_OK;
>> +
>> +    snprintf(command, sizeof(command), "%s", fw->lspci);
>> +
>> +    if (fwts_pipe_exec(command,&lspci_output) == FWTS_EXEC_ERROR) {
>> +        fwts_log_warning(fw, "Could not execute %s", command);
>> +        return FWTS_EXEC_ERROR;
>> +    }
>> +
> fwts_pipe_exec may return a NULL list (e.g. out of of memory), so we 
> should check for that.  It's a poor API :-(
>
>         if (lspci_output == NULL)
>             return FWTS_ERROR;
>
>> +    /* Get the list of pci devices and their configuration */
>> +    fwts_list_foreach(item, lspci_output) {
>> +        char *line = fwts_text_list_text(item);
>> +        char *pEnd;
>> +
>> +        device = (struct pci_device *) malloc(sizeof(*device));
>
> I try to avoid malloc in fwts and use calloc instead since it also 
> ensures the memory is zero'd.
>
>> +        if (device == NULL)
>> +            return FWTS_ERROR;
>> +
>> +        device->bus = strtol(line,&pEnd, 16);
>> +        device->dev = strtol(pEnd + 1,&pEnd, 16);
>> +        device->func = strtol(pEnd + 1,&pEnd, 16);
>> +
>> +        if (device->bus == 0&&  device->dev == 0&&  device->func == 0)
>> +            root = device;
>> +        else
>> +            cur->next = device;
>
> I suspect we are assuming that lspci will always output in device 
> order, with the root at the first item. My fear is that if this does 
> not happen the cur->next will segfault if cur is NULL. Perhaps we should
> check for that to bullet-proof the code.
Very good point, and I think I don't need to find the root device; 
instead, all I need is the first device output from lspci (just need a 
head of the for linked list). I will change the name of the variable and 
modify the code accordingly.
>
>> +
>> +        cur = device;
>> +    }
>> +
>
> I'd prefer:
>     for (cur = root; cur; cur = cur->next)
>
> instead of:
>     cur = root;
>     while (cur != NULL) {
>         ...
>         ...
>         cur = cur->next;
>     }
>
> ..several times in the code below.
>
>> +    cur = root;
>> +    while (cur != NULL) {
>> +        int reg_loc = 0, reg_val = 0;
>> +        int i;
>> +
>> +        snprintf(command, sizeof(command), "%s -s %02X:%02X.%02X -xxx",
>> +            fw->lspci, cur->bus, cur->dev, cur->func);
>> +        if (fwts_pipe_exec(command,&lspci_output) == FWTS_EXEC_ERROR) {
>> +            fwts_log_warning(fw, "Could not execute %s", command);
>> +            return FWTS_EXEC_ERROR;
>
> Should be return FWTS_ERROR so that the test framework can handle this 
> appropriately (FWTS_ERROR, FWTS_SKIP, FWTS_OK are the only return 
> codes that should be returned up to the fwts framework).
>
>> +        }
>
> Again, null list check required:
>
>         if (lspci_output == NULL)
>             return FWTS_ERROR;
>
> OK, so we're parsing something like:
>
> 00:1b.0 Audio device: Intel Corporation 82801H (ICH8 Family) HD Audio 
> Controller (rev 03)
> 00: 86 80 4b 28 06 05 10 00 03 00 03 04 10 00 00 00
> 10: 04 00 30 fc 00 00 00 00 00 00 00 00 00 00 00 00
> 20: 00 00 00 00 00 00 00 00 00 00 00 00 aa 17 4e 38
> 30: 00 00 00 00 50 00 00 00 00 00 00 00 0b 01 00 00
>
>> +        fwts_list_foreach(item, lspci_output) {
>> +            char *line = fwts_text_list_text(item);
>> +            char *pEnd;
>> +
>> +            if (line[3] == ' ') {
>> +                reg_val = strtol(line,&pEnd, 16);
>> +                for (i = 0; i<  16; i++) {
>> +                    reg_val = strtol(pEnd + 1,&pEnd, 16);
>> +                    cur->config[reg_loc] = reg_val;
>> +                    reg_loc++;
>
> Hopefully we won't have > 256 bytes in PCI config space. Maybe we 
> should guard against thsi overflowing cur->config[]
PCI configuration space will always be exact 256 bytes long (the 
addition pci extended space is exactly 4096 byte long, and it is from 
lspci -xxxx).
Assuming lspci -xxx will not be changed to display extended 
configuration space, this should be safe.
>
>> +                }
>> +            }
>> +        }
>> +        cur = cur->next;
>> +    }
>> +
>> +    /* Check aspm registers from the list of pci devices */
>> +    cur = root;
>> +    while (cur != NULL) {
>> +        struct pci_device *target;
>> +
>> +        if (cur->config[0x0E]&  0x01) {
>
> These magic numbers mean nothing to me. Any chance of #defining them?
>
>> +
>
> Again, I'd prefer:
>
>     for (target = root; target; target = target->next)
>
>> +            target = root;
>> +            while (target != NULL) {
>> +                if (target->bus == cur->config[0x19])
>> +                    break;
>> +                target = target->next;
>> +            }
>> +            if (target == NULL) {
>> +                cur = cur->next;
>> +                continue;
>> +            }
>> +
>> +            pcie_compare_rp_dev_aspm_registers(fw, cur, target);
>> +        }
>> +        cur = cur->next;
>> +    }
>> +
>> +    cur = root;
>> +    while (cur != NULL) {
>> +        device = cur->next;
>> +        free(cur);
>> +        cur = device;
>> +    }
>> +
>> +    fwts_text_list_free(lspci_output);
>> +
>> +    return ret;
>> +}
>> +
>>   static int aspm_check_configuration(fwts_framework *fw)
>>   {
>>       int ret;
>> @@ -65,8 +268,18 @@ static int 
>> aspm_check_configuration(fwts_framework *fw)
>>       return ret;
>>   }
>>
>> +static int aspm_pcie_register_configuration(fwts_framework *fw)
>> +{
>> +    int ret;
>> +
>> +    ret = pcie_check_aspm_registers(fw, LOG_LEVEL_HIGH);
>> +
>> +    return ret;
>> +}
>
> how about:
>
> {
>     return pcie_check_aspm_registers(fw, LOG_LEVEL_HIGH);
> }
>
>> +
>>   static fwts_framework_minor_test aspm_tests[] = {
>> -    { aspm_check_configuration, "PCIe ASPM configuration test." },
>> +    { aspm_check_configuration, "PCIe ASPM ACPI test." },
>> +    { aspm_pcie_register_configuration, "PCIe ASPM registers test." },
>>       { NULL, NULL }
>>   };
>>
>
> Conceptually this is great work, just needs a little more re-working. 
> Thanks for the contribution Alex!
>
> Colin
>
Thanks a lot, I will re-work the patch according to the feedback.

Best Regards,
Alex Hung





More information about the fwts-devel mailing list