On 11/4/05, Xiangfei Jia <xjianz@xxxxxxxxx> wrote: > On 11/4/05, Fawad Lateef <fawadlateef@xxxxxxxxx> wrote: > > On 11/4/05, Xiangfei Jia <xjianz@xxxxxxxxx> wrote: > > > > > > For block devices, the "blkdevs" array is used to hold all the > registered > > > block device drivers, and the maximum number of allowed block drivers is > > > 255. There is another hash table array called "bdev_hashtable", which is > > > array of struct "block_device". > > > > > > I'm quite confused with these two arrays. (1) what is the relationship > > > between the "blkdevs" and the "bdev_hashtable"?? > > > > blkdevs is an array used internally by block_dev.c (block-device > > layer) to keep track of registered block devices whereas > > bdev_hashtable array of type list-head is used to keep information > > related to block_device in a hash formed which actually hold the > > struct block_device type pointers/nodes (field bd_hash) > Why is "blkdevs" only used in block-device layer? Why not just use > "bdev_hashtable" everywhere? I just can't understand why use two different > arrays to represent the same thing twice. > I think blkdevs is used to keep the things simple, as one might just register a block device and might be doing some work internally, but no one is going to open the device or claiming the device (calling bdget) then bdev_hashtable won't contain any entry related to that device hence different memory allocations (for filling the struct block_device and addition of it in bdev_hashtable) won't be done. (Its my thinking and understanding from the code block_dev.c, not 100% sure about this) > And it seems that "bdev_hashtable" is also a local variable in > "block_dev.c" file. Here is the code: > static struct list_head bdev_hashtable[ > HASH_SIZE]; > > > > > > (2) Why a extra sruct, called "block_device" is necessary to be used? > > > > > > > struct block_device is actually there for keeping information about > > the block_devices like bd_count (how-many times block_device is > > claimed through bdget function), bd_inode is the inode having > > data/fields filled with the block_device information see bdget > > function > (http://lxr.linux.no/source/fs/block_dev.c?v=2.4.28#L331), > > bd_openers is used to keep track of how many times block_device is > > opened and so other fields too used to keep information related to > > block_device usage. Just look for struct block_device fields used in > > block_dev.c file ... > > > > I hope my explaination might clear some doubts for you ! > Are you saying that "struct block_device" acutally keeps the really info > about each block device driver, Not the "blkdevs" array? > Ya, I want to say that struct block_device is maintaining the device related information about its count, inode etc but blkdevs have only name and struct block_device_operations !! > > By the way these blkdevs and bdev_hashtable (due to their limitations) > > are removed in 2.6.xx series kernels > > Sorry, I forgot to mention that I'm using kernel 2.4.28. > And I think that keeping the same thing in two different arrays might be one of the many reasons to remove those from 2.6.xx kenel -- Fawad Lateef -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
