The problem is caused by the incorrect handling of the parent inode's
i_nlink count for the dentry to be RENAME_EXCHANGED. There are 3 cases
to consider. Assume we want to RENAME_EXCHANGE struct dentry *a and
struct dentry *b, and inode_a is pointed to by dentry_a, inode_b is
pointed to by dentry_b:
1. If inode_a is a directory, but inode_b isn't, then we must decrease
the i_nlink count of old_dir_i, and increase the i_nlink of new_dir_i.
2. If inode_a isn't a directory, but inode_b is a directory, then we
must increase the i_nlink of old_dir_i, and decrease the i_nlink count
of new_dir_i.
3. If the types of inode_a and inode_b are the same, we don't change the
i_nlink for either old_dir_i or new_dir_i.
Signed-off-by: Jing Qiu <aqiu0720@gmail.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Fix a kernel crash caused when CONFIG_FIXED_PHY used for fixed phy drivers
in phy-add-aneg-done-function patch.
Signed-off-by: Tim Harvey <tharvey@gateworks.com>
This backports fixes for setting of_node and making it possible to read
extra info from DT. This was partially fixed by:
[PATCH] leds: leds-gpio: Set of_node for created LED devices
but it didn't work during initialization.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Somewhere between 4.4 and 4.9 there was a change that made
OF_EARLYCON_DECLARE adding entries to the __earlycon_table instead of
__earlycon_of_table. We need to keep this table as well.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Fixes the following build error:
CC arch/mips/boot/compressed/decompress.o
OBJCOPY arch/mips/boot/compressed/vmlinux.bin
XZKERN arch/mips/boot/compressed/vmlinux.bin.z
OBJCOPY arch/mips/boot/compressed/piggy.o
LD vmlinuz
arch/mips/boot/compressed/decompress.o: In function `lzma2_lzma':
/tmp2/mehrtens/linux/arch/mips/boot/compressed/../../../../lib/xz/xz_dec_lzma2.c:884: undefined reference to `memmove'
make[1]: *** [vmlinuz] Error 1
make: *** [vmlinuz] Error 2
lib/decompress_unxz.c only defines its own memmove function if there is no
macro with the same name.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
When memory is tight, modules may need to be loaded into vmalloc()
space. The code then has to generate jump trampolines which enable
relocations between vmalloc space and physical address space.
The code had a bug that was freeing these trampolines even when the
module was successfully loaded.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Upstream linux commit 7523e4dc5057e "module: use a structure to
encapsulate layout." moves some of the struct module members into a
separate struct module_layout.
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
These are patches queued for 4.11. It adds support for even more hw and
removes some annoying WARN_ONCE.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
