Message ID | 20200820185110.17828-2-brgl@bgdev.pl (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | devres: provide and use devm_krealloc() | expand |
On Thu, Aug 20, 2020 at 08:51:08PM +0200, Bartosz Golaszewski wrote: > From: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > Implement the managed variant of krealloc(). This function works with > all memory allocated by devm_kmalloc() (or devres functions using it > implicitly like devm_kmemdup(), devm_kstrdup() etc.). > > Managed realloc'ed chunks can be manually released with devm_kfree(). > > Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> > Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> ... > +static struct devres *to_devres(void *data) > +{ > + return data - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > +} > + > +static size_t devres_data_size(size_t total_size) > +{ > + return total_size - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > +} I'm fine with above, but here is a side note, perhaps offsetof(struct devres, data) will be more practical (no duplication of alignment and hence slightly better maintenance)? (Note, I didn't check if it provides the correct result) Another side note: do we have existing users of these helpers?
On Fri, Aug 21, 2020 at 11:15:55AM +0300, Andy Shevchenko wrote: > On Thu, Aug 20, 2020 at 08:51:08PM +0200, Bartosz Golaszewski wrote: ... > > +static struct devres *to_devres(void *data) > > +{ > > + return data - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > > +} > > + > > +static size_t devres_data_size(size_t total_size) > > +{ > > + return total_size - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > > +} > > I'm fine with above, but here is a side note, perhaps > > offsetof(struct devres, data) > > will be more practical (no duplication of alignment and hence slightly better > maintenance)? (Note, I didn't check if it provides the correct result) And if the result is correct wouldn't be first a simple container_of() after all? > Another side note: do we have existing users of these helpers?
On Fri, Aug 21, 2020 at 10:35 AM Andy Shevchenko <andriy.shevchenko@linux.intel.com> wrote: > > On Thu, Aug 20, 2020 at 08:51:08PM +0200, Bartosz Golaszewski wrote: > > From: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > > > Implement the managed variant of krealloc(). This function works with > > all memory allocated by devm_kmalloc() (or devres functions using it > > implicitly like devm_kmemdup(), devm_kstrdup() etc.). > > > > Managed realloc'ed chunks can be manually released with devm_kfree(). > > > > Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> > > ... > > > +static struct devres *to_devres(void *data) > > +{ > > + return data - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > > +} > > + > > +static size_t devres_data_size(size_t total_size) > > +{ > > + return total_size - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > > +} > > I'm fine with above, but here is a side note, perhaps > > offsetof(struct devres, data) > > will be more practical (no duplication of alignment and hence slightly better > maintenance)? (Note, I didn't check if it provides the correct result) > Hi Andy, The data pointer in struct devres is defined as: u8 __aligned(ARCH_KMALLOC_MINALIGN) data[]; And this value (assigned the value of ARCH_DMA_MINALIGN) varies from one arch to another. I wasn't really sure if offsetof() would work for every case so I went with something very explicit. > Another side note: do we have existing users of these helpers? > Which ones? Because I assume you're not referring to the ones I'm adding in this patch. :) Bart
On Fri, Aug 21, 2020 at 10:59:19AM +0200, Bartosz Golaszewski wrote: > On Fri, Aug 21, 2020 at 10:35 AM Andy Shevchenko > <andriy.shevchenko@linux.intel.com> wrote: > > On Thu, Aug 20, 2020 at 08:51:08PM +0200, Bartosz Golaszewski wrote: > > > From: Bartosz Golaszewski <bgolaszewski@baylibre.com> ... > > > +static struct devres *to_devres(void *data) > > > +{ > > > + return data - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > > > +} > > > + > > > +static size_t devres_data_size(size_t total_size) > > > +{ > > > + return total_size - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > > > +} > > > > I'm fine with above, but here is a side note, perhaps > > > > offsetof(struct devres, data) > > > > will be more practical (no duplication of alignment and hence slightly better > > maintenance)? (Note, I didn't check if it provides the correct result) > > > > Hi Andy, > > The data pointer in struct devres is defined as: > > u8 __aligned(ARCH_KMALLOC_MINALIGN) data[]; > > And this value (assigned the value of ARCH_DMA_MINALIGN) varies from > one arch to another. I wasn't really sure if offsetof() would work for > every case so I went with something very explicit. I have checked with a small program simulating to_devres() with your variant, offsetof() and container_of(). The result is this: if MINALIGN < sizeof(long) and since struct is unpacked the offsetof(), and thus container_of(), gives correct result, while ALIGN() approach mistakenly moves pointer too back. Test results ~~~~~~~~~~~~ sizeof(devres), ALIGN(data), resulting 3 pointers against NULL followed by actual pointer of allocated struct and 3 pointers against it. % for i in 1 2 4; do gcc -O2 -Wall -DSZ=$i -o test test.c && ./test | head -n1; done szof: 24 a: 24 0xffffffffffffffe8 0xffffffffffffffef 0xffffffffffffffef, 0x55a3aa91e2a0: 0x55a3aa91e299 0x55a3aa91e2a0 0x55a3aa91e2a0 szof: 24 a: 24 0xffffffffffffffe8 0xffffffffffffffee 0xffffffffffffffee, 0x563d7b88b2a0: 0x563d7b88b29a 0x563d7b88b2a0 0x563d7b88b2a0 szof: 24 a: 24 0xffffffffffffffe8 0xffffffffffffffec 0xffffffffffffffec, 0x557d08cf82a0: 0x557d08cf829c 0x557d08cf82a0 0x557d08cf82a0 % for i in 1 2 4; do gcc -m32 -O2 -Wall -DSZ=$i -o test test.c && ./test | head -n1; done szof: 12 a: 12 0xfffffff4 0xfffffff7 0xfffffff7, 0x584301a0: 0x5843019d 0x584301a0 0x584301a0 szof: 12 a: 12 0xfffffff4 0xfffffff6 0xfffffff6, 0x57bd61a0: 0x57bd619e 0x57bd61a0 0x57bd61a0 szof: 12 a: 12 0xfffffff4 0xfffffff4 0xfffffff4, 0x56b491a0: 0x56b491a0 0x56b491a0 0x56b491a0 I think you need to change this to use container_of() and offsetof(). > > Another side note: do we have existing users of these helpers? > > Which ones? Because I assume you're not referring to the ones I'm > adding in this patch. :) Opposite, the ones you are introduced here. Meaning that we might convert existing user(s) in the separate change(s) later on.
On Fri, Aug 21, 2020 at 01:51:19PM +0300, Andy Shevchenko wrote: > On Fri, Aug 21, 2020 at 10:59:19AM +0200, Bartosz Golaszewski wrote: > > On Fri, Aug 21, 2020 at 10:35 AM Andy Shevchenko > > <andriy.shevchenko@linux.intel.com> wrote: > > > On Thu, Aug 20, 2020 at 08:51:08PM +0200, Bartosz Golaszewski wrote: ... > > > > +static struct devres *to_devres(void *data) > > > > +{ > > > > + return data - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > > > > +} > > > > + > > > > +static size_t devres_data_size(size_t total_size) > > > > +{ > > > > + return total_size - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > > > > +} > > The data pointer in struct devres is defined as: > > > > u8 __aligned(ARCH_KMALLOC_MINALIGN) data[]; > > > > And this value (assigned the value of ARCH_DMA_MINALIGN) varies from > > one arch to another. I wasn't really sure if offsetof() would work for > > every case so I went with something very explicit. > > I have checked with a small program simulating to_devres() with your variant, > offsetof() and container_of(). > > The result is this: if MINALIGN < sizeof(long) and since struct is unpacked the > offsetof(), and thus container_of(), gives correct result, while ALIGN() > approach mistakenly moves pointer too back. ... > I think you need to change this to use container_of() and offsetof(). To be clear, there is probably no real problem, except unlikely possible MINALIGN=4 on 64-bit arch, but for sake of the correctness.
On Fri, Aug 21, 2020 at 1:04 PM Andy Shevchenko <andriy.shevchenko@linux.intel.com> wrote: > > On Fri, Aug 21, 2020 at 01:51:19PM +0300, Andy Shevchenko wrote: > > On Fri, Aug 21, 2020 at 10:59:19AM +0200, Bartosz Golaszewski wrote: > > > On Fri, Aug 21, 2020 at 10:35 AM Andy Shevchenko > > > <andriy.shevchenko@linux.intel.com> wrote: > > > > On Thu, Aug 20, 2020 at 08:51:08PM +0200, Bartosz Golaszewski wrote: > > ... > > > > > > +static struct devres *to_devres(void *data) > > > > > +{ > > > > > + return data - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > > > > > +} > > > > > + > > > > > +static size_t devres_data_size(size_t total_size) > > > > > +{ > > > > > + return total_size - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); > > > > > +} > > > > The data pointer in struct devres is defined as: > > > > > > u8 __aligned(ARCH_KMALLOC_MINALIGN) data[]; > > > > > > And this value (assigned the value of ARCH_DMA_MINALIGN) varies from > > > one arch to another. I wasn't really sure if offsetof() would work for > > > every case so I went with something very explicit. > > > > I have checked with a small program simulating to_devres() with your variant, > > offsetof() and container_of(). > > > > The result is this: if MINALIGN < sizeof(long) and since struct is unpacked the > > offsetof(), and thus container_of(), gives correct result, while ALIGN() > > approach mistakenly moves pointer too back. > > ... > > > I think you need to change this to use container_of() and offsetof(). > > To be clear, there is probably no real problem, except unlikely possible > MINALIGN=4 on 64-bit arch, but for sake of the correctness. > Thanks for taking the time to check it. I'll switch to container_of() for (hopefully) the last iteration. Bart
diff --git a/Documentation/driver-api/driver-model/devres.rst b/Documentation/driver-api/driver-model/devres.rst index eaaaafc21134..f318a5c0033c 100644 --- a/Documentation/driver-api/driver-model/devres.rst +++ b/Documentation/driver-api/driver-model/devres.rst @@ -354,6 +354,7 @@ MEM devm_kmalloc() devm_kmalloc_array() devm_kmemdup() + devm_krealloc() devm_kstrdup() devm_kvasprintf() devm_kzalloc() diff --git a/drivers/base/devres.c b/drivers/base/devres.c index ed615d3b9cf1..55f3abb966ee 100644 --- a/drivers/base/devres.c +++ b/drivers/base/devres.c @@ -36,6 +36,16 @@ struct devres { u8 __aligned(ARCH_KMALLOC_MINALIGN) data[]; }; +static struct devres *to_devres(void *data) +{ + return data - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); +} + +static size_t devres_data_size(size_t total_size) +{ + return total_size - ALIGN(sizeof(struct devres), ARCH_KMALLOC_MINALIGN); +} + struct devres_group { struct devres_node node[2]; void *id; @@ -126,6 +136,14 @@ static void add_dr(struct device *dev, struct devres_node *node) list_add_tail(&node->entry, &dev->devres_head); } +static void replace_dr(struct device *dev, + struct devres_node *old, struct devres_node *new) +{ + devres_log(dev, old, "REPLACE"); + BUG_ON(!list_empty(&new->entry)); + list_replace(&old->entry, &new->entry); +} + #ifdef CONFIG_DEBUG_DEVRES void * __devres_alloc_node(dr_release_t release, size_t size, gfp_t gfp, int nid, const char *name) @@ -837,6 +855,102 @@ void *devm_kmalloc(struct device *dev, size_t size, gfp_t gfp) } EXPORT_SYMBOL_GPL(devm_kmalloc); +/** + * devm_krealloc - Resource-managed krealloc() + * @dev: Device to re-allocate memory for + * @ptr: Pointer to the memory chunk to re-allocate + * @new_size: New allocation size + * @gfp: Allocation gfp flags + * + * Managed krealloc(). Resizes the memory chunk allocated with devm_kmalloc(). + * Behaves similarly to regular krealloc(): if @ptr is NULL or ZERO_SIZE_PTR, + * it's the equivalent of devm_kmalloc(). If new_size is zero, it frees the + * previously allocated memory and returns ZERO_SIZE_PTR. This function doesn't + * change the order in which the release callback for the re-alloc'ed devres + * will be called (except when falling back to devm_kmalloc() or when freeing + * resources when new_size is zero). The contents of the memory are preserved + * up to the lesser of new and old sizes. + */ +void *devm_krealloc(struct device *dev, void *ptr, size_t new_size, gfp_t gfp) +{ + size_t total_new_size, total_old_size; + struct devres *old_dr, *new_dr; + unsigned long flags; + + if (unlikely(!new_size)) { + devm_kfree(dev, ptr); + return ZERO_SIZE_PTR; + } + + if (unlikely(ZERO_OR_NULL_PTR(ptr))) + return devm_kmalloc(dev, new_size, gfp); + + if (WARN_ON(is_kernel_rodata((unsigned long)ptr))) + /* + * We cannot reliably realloc a const string returned by + * devm_kstrdup_const(). + */ + return NULL; + + if (!check_dr_size(new_size, &total_new_size)) + return NULL; + + total_old_size = ksize(to_devres(ptr)); + if (total_old_size == 0) { + WARN(1, "Pointer doesn't point to dynamically allocated memory."); + return NULL; + } + + /* + * If new size is smaller or equal to the actual number of bytes + * allocated previously - just return the same pointer. + */ + if (total_new_size <= total_old_size) + return ptr; + + /* + * Otherwise: allocate new, larger chunk. We need to allocate before + * taking the lock as most probably the caller uses GFP_KERNEL. + */ + new_dr = alloc_dr(devm_kmalloc_release, + total_new_size, gfp, dev_to_node(dev)); + if (!new_dr) + return NULL; + + /* + * The spinlock protects the linked list against concurrent + * modifications but not the resource itself. + */ + spin_lock_irqsave(&dev->devres_lock, flags); + + old_dr = find_dr(dev, devm_kmalloc_release, devm_kmalloc_match, ptr); + if (!old_dr) { + spin_unlock_irqrestore(&dev->devres_lock, flags); + devres_free(new_dr); + WARN(1, "Memory chunk not managed or managed by a different device."); + return NULL; + } + + replace_dr(dev, &old_dr->node, &new_dr->node); + + spin_unlock_irqrestore(&dev->devres_lock, flags); + + /* + * We can copy the memory contents after releasing the lock as we're + * no longer modyfing the list links. + */ + memcpy(new_dr->data, old_dr->data, devres_data_size(total_old_size)); + /* + * Same for releasing the old devres - it's now been removed from the + * list. This is also the reason why we must not use devm_kfree() - the + * links are no longer valid. + */ + kfree(old_dr); + + return new_dr->data; +} +EXPORT_SYMBOL_GPL(devm_krealloc); + /** * devm_kstrdup - Allocate resource managed space and * copy an existing string into that. diff --git a/include/linux/device.h b/include/linux/device.h index ca18da4768e3..5da7d5f0a7ff 100644 --- a/include/linux/device.h +++ b/include/linux/device.h @@ -206,6 +206,8 @@ int devres_release_group(struct device *dev, void *id); /* managed devm_k.alloc/kfree for device drivers */ void *devm_kmalloc(struct device *dev, size_t size, gfp_t gfp) __malloc; +void *devm_krealloc(struct device *dev, void *ptr, size_t size, + gfp_t gfp) __must_check; __printf(3, 0) char *devm_kvasprintf(struct device *dev, gfp_t gfp, const char *fmt, va_list ap) __malloc; __printf(3, 4) char *devm_kasprintf(struct device *dev, gfp_t gfp,