x86 boot : export boot_params via sysfs (forward to Greg)
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 23 lines · Ying
Quoted 13 lines · Greg KH
Quoted 13 lines
(quoted diff) +13
-0+static struct attribute *boot_params_attrs[] = { + &boot_params_version_attr.attr, + NULL +}; + +static struct attribute_group boot_params_attr_group = { + .attrs = boot_params_attrs, +}; + +static ssize_t boot_params_data_read(struct kobject *kobj, + struct bin_attribute *bin_attr, + char *buf, loff_t off, size_t count) +{Quoted 8 lines
Review
Quoted 3 lines
(quoted diff) +3
-0+ memcpy(buf, (void *)&boot_params + off, count); + return count; +}Why is the whole structure being exported to userspace directly? Traditionally we only use the binary files for blobs that are not interpreted by the kernel in any way. Is that what this structure really is?
Quoted 4 lines
The struct
boot_paramsis not the case. But it is too complex (such as the structedd_infoin it). It is very complex to output every fields include the first level fields until the primary data type. And Peter Anvin think it is OK to output it as a big binary file.
Well, I respectively disagree. sysfs is NOT for exporting various binary kernel structures to userspace directly. Again, the binary files in sysfs are for chunks of memory that are PASS-THROUGH from hardware to userspace, with no kernel intervention at all.
If you really need such a thing, use debugfs, as the only rule for debugfs is that there is no rules :)
So I do not agree with this change at this time, and do not want to see it applied to the tree.
If you really need this kind of information exported to userspace through sysfs, then expand it all, in a one-text-value-per-file format. I would have no objection to that.
thanks,
greg k-h
Quoted 6 lines · Greg Kroah-Hartman
Well, I respectively disagree. sysfs is NOT for exporting various binary kernel structures to userspace directly. Again, the binary files in sysfs are for chunks of memory that are PASS-THROUGH from hardware to userspace, with no kernel intervention at all.
If you really need such a thing, use debugfs, as the only rule for debugfs is that there is no rules :)
Whilst on the subject, why wasn't /sys/slab done in debugfs ? The one-value-per-file thing has gone taken to ridiculous extremes there.
Having 3641 sysfs files that most people never use permanently taking up memory seems to be a massive waste of resources.
Nearly a third of all the sysfs files I have on my system belong to that subtree, which is just.. wow.
Dave
Signature · 1 line
http://www.codemonkey.org.uk
Quoted 11 lines · Dave Jones
Quoted 6 lines · Greg Kroah-Hartman
Well, I respectively disagree. sysfs is NOT for exporting various binary kernel structures to userspace directly. Again, the binary files in sysfs are for chunks of memory that are PASS-THROUGH from hardware to userspace, with no kernel intervention at all.
If you really need such a thing, use debugfs, as the only rule for debugfs is that there is no rules :)Whilst on the subject, why wasn't
/sys/slabdone in debugfs ? The one-value-per-file thing has gone taken to ridiculous extremes there.
Heh, I think most of those files only show up if you have CONFIG_SLUB_DEBUG enabled, and it's easier to add them to the original sysfs entries instead of creating new debugfs ones.
Quoted 2 lines
Having 3641 sysfs files that most people never use permanently taking up memory seems to be a massive waste of resources.
sysfs files have their dentries and inodes pushed out of memory if they are never accessed, so no huge ammounts of memory are used here.
Otherwise we would never be able to support 20,000 block devices on a 31bit s390 with 128Mb of ram :)
thanks,
greg k-h