|
From: Philippe E. <ph...@wa...> - 2003-04-17 00:03:51
Attachments:
syscall-253.diff
|
hi, this patch, against current cvs, implement syscall lookup_dcookie added in 2.5 It allows to associate an uid with an existing dentry and to query the corresponding filename. This is to provide fast transfer between kernel and user space of filename through 64 bits uid. It is used currently only by profiler but its intended to be usefull for tracing driver of any sort which need to associate event to filename and pass them to a daemon. The post_mem_write use arg4 (sz) as size since the syscall doesn't return the string length on success but zero. I tested it with oprofile, please consider to apply. regards, Phil |
|
From: Nicholas N. <nj...@ca...> - 2003-05-01 10:59:24
|
On Thu, 17 Apr 2003, Philippe Elie wrote: > this patch, against current cvs, implement syscall > lookup_dcookie added in 2.5 > > It allows to associate an uid with an existing dentry > and to query the corresponding filename. This is to > provide fast transfer between kernel and user space of > filename through 64 bits uid. > > It is used currently only by profiler but its intended > to be usefull for tracing driver of any sort which need > to associate event to filename and pass them to a daemon. > > The post_mem_write use arg4 (sz) as size since the syscall > doesn't return the string length on success but zero. ftp.kernel.org/pub/linux/docs/manpages/man-pages-1.56.tar.gz says that the return value is the length of the string, is it incorrect? N |
|
From: Philippe E. <ph...@wa...> - 2003-05-01 18:13:45
Attachments:
lookup_dcookie.vg.diff
|
Nicholas Nethercote wrote: > On Thu, 17 Apr 2003, Philippe Elie wrote: > > >>this patch, against current cvs, implement syscall >>lookup_dcookie added in 2.5 >> >>It allows to associate an uid with an existing dentry >>and to query the corresponding filename. This is to >>provide fast transfer between kernel and user space of >>filename through 64 bits uid. >> >>It is used currently only by profiler but its intended >>to be usefull for tracing driver of any sort which need >>to associate event to filename and pass them to a daemon. >> >>The post_mem_write use arg4 (sz) as size since the syscall >>doesn't return the string length on success but zero. > > > ftp.kernel.org/pub/linux/docs/manpages/man-pages-1.56.tar.gz says that the > return value is the length of the string, is it incorrect? > man page is right, my comment was about an older implementation, looking at getcwd syscall there is exactly the same problem, getcwd return the length written but valgrind use VG_TRACK( post_mem_write, arg1, arg2 ); I attach the updated patch, choose the once you prefer, the new patch is better but not consistent with sys_getcwd behavior. regards, Phil |