Le 19/08/2024 à 18:51, Philipp Stanner a écrit :
solidrun utilizes pcim_iomap_regions(), which has been deprecated by the PCI subsystem in commit e354bb84a4c1 ("PCI: Deprecate pcim_iomap_table(), pcim_iomap_regions_request_all()"), among other things because it forces usage of quite a complicated bitmask mechanism. The bitmask handling code can entirely be removed by replacing pcim_iomap_regions() and pcim_iomap_table().Replace pcim_iomap_regions() and pcim_iomap_table() with pci_iomap_region(). Signed-off-by: Philipp Stanner <pstan...@redhat.com> --- drivers/vdpa/solidrun/snet_main.c | 47 +++++++++++-------------------- 1 file changed, 16 insertions(+), 31 deletions(-) diff --git a/drivers/vdpa/solidrun/snet_main.c b/drivers/vdpa/solidrun/snet_main.c index 99428a04068d..abf027ca35e1 100644 --- a/drivers/vdpa/solidrun/snet_main.c +++ b/drivers/vdpa/solidrun/snet_main.c @@ -556,33 +556,24 @@ static const struct vdpa_config_ops snet_config_ops = { static int psnet_open_pf_bar(struct pci_dev *pdev, struct psnet *psnet) { char name[50]; - int ret, i, mask = 0; + int i; + + snprintf(name, sizeof(name), "psnet[%s]-bars", pci_name(pdev)); + /* We don't know which BAR will be used to communicate.. * We will map every bar with len > 0. * * Later, we will discover the BAR and unmap all other BARs. */ for (i = 0; i < PCI_STD_NUM_BARS; i++) { - if (pci_resource_len(pdev, i)) - mask |= (1 << i); - } - - /* No BAR can be used.. */ - if (!mask) { - SNET_ERR(pdev, "Failed to find a PCI BAR\n"); - return -ENODEV; - } - - snprintf(name, sizeof(name), "psnet[%s]-bars", pci_name(pdev)); - ret = pcim_iomap_regions(pdev, mask, name); - if (ret) { - SNET_ERR(pdev, "Failed to request and map PCI BARs\n"); - return ret; - } + if (pci_resource_len(pdev, i)) { + psnet->bars[i] = pcim_iomap_region(pdev, i, name);
Hi, Unrelated to the patch, but is is safe to have 'name' be on the stack? pcim_iomap_region() --> __pcim_request_region() --> __pcim_request_region_range() --> request_region() or __request_mem_region() --> __request_region() --> __request_region_locked() --> res->name = name; So an address on the stack ends in the 'name' field of a "struct resource". According to a few grep, it looks really unusual. I don't know if it is used, but it looks strange to me. If it is an issue, it was apparently already there before this patch.
+ if (IS_ERR(psnet->bars[i])) { + SNET_ERR(pdev, "Failed to request and map PCI BARs\n"); + return PTR_ERR(psnet->bars[i]); + } + }- for (i = 0; i < PCI_STD_NUM_BARS; i++) {- if (mask & (1 << i)) - psnet->bars[i] = pcim_iomap_table(pdev)[i]; }return 0;@@ -591,18 +582,15 @@ static int psnet_open_pf_bar(struct pci_dev *pdev, struct psnet *psnet) static int snet_open_vf_bar(struct pci_dev *pdev, struct snet *snet) { char name[50]; - int ret;snprintf(name, sizeof(name), "snet[%s]-bar", pci_name(pdev));/* Request and map BAR */ - ret = pcim_iomap_regions(pdev, BIT(snet->psnet->cfg.vf_bar), name); - if (ret) { + snet->bar = pcim_iomap_region(pdev, snet->psnet->cfg.vf_bar, name);
Same Just my 2c. CJ
+ if (IS_ERR(snet->bar)) { SNET_ERR(pdev, "Failed to request and map PCI BAR for a VF\n"); - return ret; + return PTR_ERR(snet->bar); }- snet->bar = pcim_iomap_table(pdev)[snet->psnet->cfg.vf_bar];- return 0; }@@ -650,15 +638,12 @@ static int psnet_detect_bar(struct psnet *psnet, u32 off) static void psnet_unmap_unused_bars(struct pci_dev *pdev, struct psnet *psnet){ - int i, mask = 0; + int i;for (i = 0; i < PCI_STD_NUM_BARS; i++) {if (psnet->bars[i] && i != psnet->barno) - mask |= (1 << i); + pcim_iounmap_region(pdev, i); } - - if (mask) - pcim_iounmap_regions(pdev, mask); }/* Read SNET config from PCI BAR */
