Return-Path: To: Bruno Vidal Cc: "parisc-linux@lists.parisc-linux.org" Subject: Re: [parisc-linux] How retrieving HPA/SPA with the device node In-Reply-To: Message from Bruno Vidal of "Tue, 02 Apr 2002 16:08:16 +0200." <3CA9BB50.1FCEBDF4@admin.france.hp.com> References: <3CA9BB50.1FCEBDF4@admin.france.hp.com> Date: Tue, 02 Apr 2002 09:44:17 -0700 From: Grant Grundler Message-Id: <20020402164417.6C6B6482A@dsl2.external.hp.com> Sender: parisc-linux-admin@lists.parisc-linux.org Errors-To: parisc-linux-admin@lists.parisc-linux.org List-Help: List-Post: List-Subscribe: , List-Id: parisc-linux developers list List-Unsubscribe: , List-Archive: Bruno Vidal wrote: > The pointer of the disk (/dev/sd??) is the node. Example: > /dev/sda2 ->> sda2 -> 0x802 -> major:0x8 -> driver "sd" with minor:0x02 > Now how can I link 0x802 to hardware path "2/0/1" > So where is the link between both ????? Do you have any idea ????? Well, there is a link. It's just not as obvious as the iotree. drivers/scsi/sd.c:sd_open() finds the device: target = DEVICE_NR(inode->i_rdev); if (target >= sd_template.dev_max || !rscsi_disks[target].device) return -ENXIO; /* No such device */ And sd.h has the definition for Scsi_Disk: typedef struct scsi_disk { unsigned capacity; /* size in blocks */ Scsi_Device *device; unsigned char ready; /* flag ready for FLOPTICAL */ unsigned char write_prot; /* flag write_protect for rmvable dev */ unsigned char sector_bit_size; /* sector_size = 2 to the bit size power */ unsigned char sector_bit_shift; /* power of 2 sectors per FS block */ unsigned has_part_table:1; /* has partition table */ } Scsi_Disk; Scsi_Device is *private*. It contains the ptr to Scsi_Host (defined in hosts.h), an instance of a SCSI controller driver. And finally, Scsi_Host has a pointer to pci_dev. Since several of the data structures are intended to be private, I suggest adding a function to sd.c (or scsi.c) that finds and returns the pci_dev pointer: struct pci_dev *sd_get_pci_dev(struct inode *inode) { int target = DEVICE_NR(inode->i_rdev); Scsi_Device scsidev = rscsi_disks[target].device; if (target >= sd_template.dev_max || !scsidev) return NULL; /* No such device */ return (scsi_dev->host->pci_dev); } It would be nice if SCSI were part of the IO tree but I can understand reasons for not doing it. In particular, SCSI over FC doesn't lend itself to an IO Tree arrangement since it's really a "net". Put a SAN, additional FC controllers from the same host, and a few thousand LUNs out there and the IO tree can become an ugly mess to. hth, grant ps. I had no clue how this worked until I tracked this down starting at sd_open().